1

Clitic placement in South Slavic*

� � � � � � � � � � � �

Abstract. The paper examines cliti c placement and the nature of clit ic clustering in Serbo-Croatian, Bulgarian, and Macedonian. It is argued that Serbo-Croatian cliti cs do not cluster syntactically; they are located in different projections in the syntax. The order of cliti cs within the cliti c cluster is argued to follow from the hierarchical arrangement of projections in which they are located. The paper also provides a principled account of the idiosyncratic behavior of the auxili ary cliti c je, which in contrast to other auxili ary cliti cs follows pronominal cliti cs. In contrast to Serbo-Croatian cliti cs, Bulgarian and Macedonian cliti cs are argued to cluster in the same head position in the final syntactic representation. The cluster is formed through successive cyclic leftward adjunctions of cliti cs to the verb, in accordance with the LCA. Following Chomsky’s (1994) suggestion that cliti cs are ambiguous head/phrasal elements, it is argued that cliti cs do not branch, hence cannot take complements. This claim leads to a new proposal concerning the structural representation of several cliti c forms.

This paper examines cliti c placement and the nature of cliti c clustering in the South Slavic

languages. On the more theoretical side, the paper addresses the question of whether PF can

affect word order. It also makes a proposal concerning the structural representation of cliti cs

which is meant to hold crosslinguistically. In section 1 I examine the cliti c system of Serbo-

Croatian (SC), a second position cliti c language. In section 2 I turn to Bulgarian and

Macedonian, whose cliti cs are traditionally considered to be verbal.

*Parts of this article were presented at the Workshop on Cliti c Phenomena in English and other

European Languages (Debrecen), University of Novi Sad, 5th Central European Summer School

(Debrecen), 3rd Conference on Formal Approaches to South Slavic and Balkan Languages at

University of Plovdiv (with Steven Franks), Cornell University, CUNY, Université de Paris 8,

University of Maryland, Workshop on Slavic Pronominal Cliti cs at ZAS, Berlin, Workshop on

the Syntax of South Slavic Languages at University of Venice, Acme Balkanica Conference at

Concordia University, Conference on the Balkan Sprachbund Properties at University of Leiden,

Universidade Federal do Rio de Janeiro, Universidade Estadual de Campinas, Lund University,

and University of Connecticut. I thank the audiences in all of these settings for thought-

provoking questions. For valuable comments, special thanks are due to Klaus Abels, Wayles � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � ! � � " " � � � � � � � # � � � $ %

2

1. Serbo-Croatian clitics

The phenomenon of second position cliti cization in SC is ill ustrated by (1a-d). Locating cliti cs in

any other position or splitting the cliti c cluster in (1) would lead to ungrammaticali ty. (Cliti cs are

given in italics.)

(1) a. Mi smo mu je & ' ( ) * + , - . / . 0 1 2 ( 3

we are him.dat her.acc introduced yesterday

‘We introduced her to him yesterday.’

b. Zašto smo mu je 4 5 6 7 8 9 : ; < = < > ? @ 6 A

why are him.dat her.acc introduced yesterday

‘Why did we introduce her to him yesterday?’

c. Ona tvrdi da smo mu je B < 4 5 6 7 8 9 : ; < = < > ? @ 6 C she claims that are him.dat her.acc we introduced yesterday

‘She claims that we introduced her to him yesterday.’

d. Predstavili smo mu je > ? @ 6 C introduced are him.dat her.acc yesterday

‘We introduced her to him yesterday.’

Second position cliti cization in SC has recently attracted a great deal of attention (see Anderson D E E F G H I J J I K D E L F G D E L M G H N I O P Q R J S T K U I V R J N W X Y Z [ [ [ G H N \ P N W X Y G D E E ] G D E E M R G Z [ [ [ R G Z [ [ D R G^ _ ` a _ b c d e f g h i e f a j k l l k m ^ i _ n f o p q r s m p q r t m p q q u m v e c f a p q q w m p q q q m x e b e i e f g y c z g o i1994, Dimitrova-Vulchanova 1995, Embick and Izvorski 1997, Franks 1997, 1998, 1999, 2000a,

Franks and King 2000, Franks and Progovac 1994, Halpern 1995, Hock 1992, King 1996, Law { | | } ~ � � � � � � } � � � ~ � � � � � � � � } � � � ~ � � � � � � � � } � � � ~ } � � � ~ } � � � ~ { | | | ~ � � � � � � � � �- � � � � � 1988, � � � � � � �   ¡ ¢ £ � � � ¤ � � £ ¥ ¡ ¢ ¦ § � � � ¨ � © ª « ¬ ¦ ­ ¡ � � � ¨ � © ¦ ® ¡ ¯ ° ± £   � ² � � � ³ ° � ¥ � � � � � � ´ £ µ � ² � � � � � ¶ · · · �

¸ � ¹ º ¡ ¢ ° ± º » °   ° ¢ � � �4a,b, 1997, Zec and Inkelas 1990). Most recent work on the topic focuses

on cliti c placement and the nature of the second position effect. These issues are often considered

to be related, especially in the syntactic approaches to the second position effect. However, in ¼ ½ ¾ ¿ ½ À Á Â Ã Ä Å Å Å Æ Ç Ä Å Å È Æ É Ç Ê Æ Ë Ì Í Î Ï Ð Æ Ï Ï Ð Î Ï Ñ ½ Á Ò Ò Í Î Ò Æ Ë Î Ó Æ Ë Ì Î Ó Ô Á Õ Ö Î × Î Õ Ö Î Õ Ï Ç Ø Ó Á Ï Á Ø × Ó Æ Ø Î Ù Î Õ Ïbeing accomplished entirely in the syntax and the second position effect being a phonological

effect. In this paper I will concentrate on the question of cliti c placement, mostly ignoring the

second position effect. However, a few words are in order on the nature of the second position

effect so that we can control for it during the discussion of cliti c placement.

3

1.1 The second position effect

The traditional statement that SC cliti cs are second within their clause is actually incorrect. It is

well -known that certain elements, such as appositives, fronted heavy constituents, and

parentheticals, can delay cliti c placement, which results in cliti cs occurring further than the

second position of their clause. This is shown by (2)-(4), where the cliti cs occur in the third and

the fourth position of their clause. (For discussion of delayed cliti c placement, see Bennett 1986, Ú Û Ü Ý Û Þ ß à á â â ã ä å æ æ æ ç ä å æ æ á ç ä Ú è Û é ê ë á â ì í ä á â ì ã ä î ç Þ ç è ç ê ï ð ß ñ ï ë è á â â í ä ò è ç ê Ý ó á â â ô ä ò è ç ê Ý óõ ö ÷ ø ù ö ú û ü ü ü ý þ õ ÿ � � � ö � � � � ý � � � � � � � � ý � � � ú � � õ � � � � ý � õ ÷ õ ö � � ù � - ø � � ù � � � 88, 1996, � � � � � � � � � � � � � � � � � � � � � ! " # $ � � ! " # % " & � ' ! ( � � � ) � ! � � " * � � � � + ( , -

. / , � ! 0 � � + � � 0 � � + � 1 � � � � ( + � ' !se samo Milena. 2 3 4 5 6 7 4 8 9 6 7 4 9 : ; 3 < = 7 4 > 7 ? @ : A ? B C 3 ? 7 A 8 D E F G H I J G K L I J G L M N F O P M Q R S T F R J Q K U J G V W

X Y Z [ Q K \ F ] K P kao što rekoh, oni ^ _ ` a b c d e f g h i

means that as said they will tomorrow arrive

‘ It means that, as I said, they will arrive tomorrow.’ j k l m n o p q r s n t n t n o r u v w n x n sam ti sladoled.

I your mother promised am you.dat ice cream

‘ I, your mother, promised you an ice cream.’

y z { | z } ~ � } � � � � { � � { � � � � � � � � � � � � � � � � � � � � � � � � { � � �- � { � � �

(1988), the distribution of SC

second position cliti cs, il lustrated above, can be stated in very simple prosodic terms:

(5) SC cliti cs occur in the second position of their intonational phrase.

Nespor and Vogel (1982, 1986), Selkirk (1986), and Hayes (1989), among others, have proposed

a hierarchical theory of the prosodic structure, which is determined by, but does not completely

correspond to, the syntactic structure of the sentence. One of the units of this prosodic structure

is intonational phrase (I-phrase). Following standard assumptions, I assume that unless

interrupted by a special element that forms a separate intonation domain, each clause is mapped

to a single I-phrase. More precisely, the left edge of a CP corresponds to an I-phrase boundary.

Certain elements, such as appositives, parentheticals, and heavy fronted constituents, are special

in that they form separate I-phrases, evidence for which is provided by the fact that they are

followed by pauses. (An I-phrase thus does not always correspond to a CP.) Under the most

natural pronunciation, cliti c second constructions such as (6) then contain only one I-phrase.

4

(6) Zaspao je Ivan.

fallen-asleep is Ivan

‘ Ivan fell asleep.’

In (2)-(4), on the other hand, the relevant clauses are parsed into more than one I-phrase, since

the appositive in (4), the fronted heavy constituent in (2), and the parenthetical in (3) form

separate I-phrases. This means that a new I-phrase starts after these elements. Note that the

elements in question are obligatorily followed by a pause, an indication of an I-phrase boundary. � � � � � � � � � � � � � � � � � � � � � � �   � � � � ¡ � ¡ � � ¢ � £ £ � � ¡ ¤ ¢ � � ¥ �   � � � � � ¦-§ ¨ © ª « ¬ ­ ® ® ¯ ¬ ­ ­ ° ± ² ³ ª ´ µ ¶ · ¸ ª ¹ ¯ ª ·

is clear that the cliti cs are located in the second position of their I-phrase in (2)-(4). When we

attempt to place a cliti c in the third position of its I-phrase, we get an ungrammatical sentence, as

indicated by (7), which contains only one I-phrase, namely the whole clause.1 The constructions

in (9) are also ungrammatical because, in contrast to (3)-(4), they run afoul of (5).

(7) *Petra srela je samo Milena.

Petar.acc met is only Milena.nom

‘Petar, only Milena met.’

(8) cf. Petra je srela samo Milena. º » ¼ ½ ¾ ¿ À ½ Á Â Ã Äala sam ti sladoled.

I promised am you.dat ice cream  ¾ ¿ Å Æ ½ Ç È É ½ Á Æ È Ê Ë Ì Í Î Ï ½ É Á Ä È ¾

means that they will t omorrow arrive

(10) a. cf. Ja sam tiÁ Â Ã Ä ½ Ð ½ Ì Ð ½ É Á Ð Ã É ¾

 ¾ Ñ Ò ¾ Å Æ ½ Ç È da Ê Ë Ì Í Î Ï ½ É Á Ä È ¾

To summarize, the correct descriptive generalization concerning the distribution of SC

second position cliti cs is not that they are second within their clause, but within their I-phrase,

which strongly indicates that the second position effect is phonological in nature.

An interesting confirmation of (5) È Ì Ó Ï Á Ô È É Ã É Â Õ Î Ö Ã Ò Á Ð Ð Á × È Æ Ø É ½ Î ½ Ò Ï Á Ù Ú Á Û Ü Á Ô È Ä

(2001a).

1As noted by Browne (1975), even moved constituents that are not heavy can delay cliti c

placement as long as they bear heavy contrastive stress and are followed by a pause, which

indicates that the relevant elements are forming separate I-phrases. I disregard this possibili ty

here.

5

(11) *Ko koga je poljubio?

who whom is kissed

‘Who kissed who?’ Ý Þ ß à á â ã ä å æ ã ç ä è é ê é ã ä ëje knjigu kupio?

which man which is book bought

‘Which man bought which book?’

The ungrammaticali ty of (11) is not surprising. Given Rudin’s (1988) claim that fronted wh-

phrases in SC do not form a constituent, (11) conforms to the phonological statement of the

second position effect: the cliti c is not located in the second position of its I-phrase (assuming a

straightforward mapping between syntactic and prosodic constituents). Interestingly, as shown in

(12), such constructions become better with heavier wh-phrases.2 The first wh-phrase in (12) is

followed by a pause, which I take to be an indication of an I-phrase boundary. As a result, the

cliti c is located in the second position of its I-phrase in (12). The prosodic statement of the

second position effect in (5) readily captures the contrast between (11) and (12). On the other

hand, it is diff icult to see how the contrast can be accounted for under a purely syntactic account

since all proposed analyses of multiple wh-fronting constructions assign (11) and (12) the same

syntactic structure.

ì í î ã ï é ã ç å ð Ý ß ñ ñ ñ ò ê ß ñ ñ Þ ò à ì ó å ç è ò í ò ô ô ã ë í õ ã ö õ ÷ è ø è ù ô ú å û õ å ç è ó è í è ú ò ü å ý ò õ å ã í å í (5). The

upshot of the analysis is that as a result of their PF lexical properties, SC cliti cs must encliti cize

to a constituent that is right-adjacent to an I-phrase boundary. As a result, they must be second

within their I-phrase. The analysis forces phonological clustering of I-phrase-mate cliti cs, but not

clause-mate cliti cs. However, it does not force syntactic clustering of cliti cs in the sense that it

does not force cliti cs necessarily to occur in the same head position (see section 1.3. for

discusså ã í ã ö ù þ í õ ò ô õ å ô û ü ò ô è ÿ è í õ ã ö � � ô ü å õ å ô ù à � � ÷ ë ù ê ë í ø è ú õ ÷ è ò í ò ü þ ù å ù û ú è ù è í õ è ø å í î ã ï é ã ç å ð

(2000a, 2001a), (13) is ruled out in PF because prosodic properties of the pronominal cliti c are

not satisfied. In other words, the pronominal cliti c is not located in the second position of its I-

phrase, in violation of (5).3

2The relevance of this type of construction was pointed out to me by Steven Franks (personal

communication).

3� ù ø å ù ô ë ù ù è ø å í î ã ï é ã ç å ð Ý ß ñ ñ ñ ò ê ß ñ ñ Þ ò à êda-clauses are parsed as separate I-phrases, which

means that an I-phrase boundary precedes da. As a result, the auxili ary cliti c satisfies (5).

6

(13) ...*da su � � � � ga istukli .

that are yesterday him beaten

‘ that they beat him yesterday.’

It is well -known that Slovenian differs from SC in that its cli tics can be either encliti cs or

procliti cs (see Bennett 1986, Bo � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � ! � " � � � �# $ $ % & ' ( ) * + , ) * - . / * 0 1 2 2 2 & 3 4 5 - 6 * ) * - 7 / 5 4 8 6 9 / : ; < 6 = = ) ( - 1 2 2 2 & > ( 6 ? * / + # $ % @-1984, Priestly

1993, and Toporiši A # $ % B & ) C 4 * 0 4 D < 6 ( , E F G , - / , H I , , 6 - / * J 4 ? + 4 9 / : K 1 2 2 # ) E & D < 6 / ( < 4 , D , D / 5 5 C I , D

be adjacent to an I-phrase boundary. The difference between SC and Slovenian is that in

Slovenian, cliti cs can either procliti cize or encliti cize to the host. As a result, prosodically,

nothing prevents breaking of a cliti c cluster in Slovenian by an element that is itself adjacent to

an I-L M N O P Q R S T U V O N W X Y P S R P Q N Z Q V [ U \ S ] ^ S Z [ _ ` a b b c O d c e a f g P S h Q i S U P j N T i j [ S U P S k j M [ P j W L Q O N Qindeed acceptable in Slovenian. (A possible context for (14) would be a question expressing the

speaker’s doubt about yesterday.)

(14) ?So Z l Q N O m ga pretepli?

are yesterday him beaten

‘They beat him yesterday?’

This confirms the relevance of prosodic requirements to cliti c clustering in the languages in

question.

I conclude here the discussion of the second position effect and turn to cliti c placement.

During the discussion of the syntax of cliti c placement we will need to control for the second

position effect. More precisely, we will need to make sure that (5) is obeyed. A construction that

violates (5) will be ruled out in PF independently of whether syntactic requirements of its cliti cs

are satisfied.

1.2 Prosodic Inversion

Before discussing the details of syntactic placement of SC cliti cs I will examine an important

argument for PF movement based on SC cliti cization.

Halpern (1995) proposes that in certain well -defined configurations SC cliti cs undergo a

PF movement operation. Halpern argues that if a SC cliti c is located sentence-initially in the

output of the syntax, it will move in the phonology looking for an appropriate host. The

underlying assumption here is that SC second position cli tics have a lexical requirement that

forces them to encliti cize to a stressed element. Cliti cs are allowed to move in PF in order to

7

satisfy this requirement. Given the well -defined motivation for PF movement, the movement

ends up being very local--it places the cliti c in a position immediately following the first stressed

word. Halpern (1995) calls the operation responsible for moving cliti cs in PF Prosodic Inversion

(PI). He considers it to be a last-resort operation that affects cliti cs only if their prosodic

requirements are not satisfied and moves them only the minimal distance necessary to satisfy the

requirements.4

Halpern proposes PI to account for the traditional observation (see Browne 1974 and

Comrie 1981) that SC cliti cs can be located either after the first phrase of their sentence (1P

environment), as in (15), or after the first word (1W environment), as in (16), where a cliti c

appears to break up a phrasal constituent:

n o p q r s t u s v w x y zsu vidjeli .

that man are seen

‘They saw that man.’

(16) Tog su

u s v w x y z v { | w x } {~

It is standardly assumed that in (15), where a whole phrase precedes the cliti c, syntactic

movement can provide a host for the cliti c. Halpern argues that in 1W environments such as (16)

PI provides a host for the cliti c. According to Halpern, the cliti c is sentence initial in the output

of the syntax. (Halpern actually does not provide independent evidence for this claim.) PI then

takes place in the phonology, placing the cliti c after the first stressed word, namely tog.

4The definition of PI from Halpern (1995:63) is given in (i).

� � � � � � � � � � � � � � � � � � � � � � � � � � � � �� � � � �

� �� � � �

� � � � � ��

� � � � � � � � � � � � �   ¡ ¢ � £ � ¤ ¥ ¦ § �   § ¨ � © ª §(respectively right),

¢« ¬

ª § � © ¤ ©   ¨ ¢ �­ ® ­ ¯ ° ± ² ³ ´ µ ¶ · ° ¸ ± ¹ º ¶ ³ ´ ¹ » ¼ ½ ´ ¯ ½ ´ µ µ ¾ ¿ º ¹ ¯ º ´ ¯ ¹ » » ¾ ´ ± ± ¶ · ´ ¹ º ¶ » ¾ º ° º ½ ¶ » ¶ ¸ º

(right) of X, then adjoin X to the right (left) of Y. À Á ¶ » µ ¶ ¹ º º ¹ ¯ ½ Â º ° º ½ ¶ ³ ´ Ã ½ º Ä » ¶ ¸ º Å ¶ · Ã ¶ ° ¸ º ½ ¶ Æ ¯ ° ± ² ° µ ¶ · ° ¸ µ ¾ ¿ º ¹ ¯ º ´ ¯ ± ¹ º ¶ ³ ´ ¹ » immediately to its right (left).

Similar operations were proposed for other languages by Marantz (1988, 1989), Sproat (1988),

Sadock (1991), and Taylor (1990). For Prosodic Inversion analyses of SC, see also Embick and

Izvorski (1997), King (1996), Percus (1993), and Schütze (1994).

8

(17) a. Syntax: su Ç È É Ê È Ë Ì Í Î Ï Ë Ð Ñ Ì Í Ò Ð Ó b. PF: Tog su Ê ovjeka vidjeli .

PI may be the strongest case ever made for PF movement. A number of constructions have been

recently suggested to involve PF movement, for example, traditional rightward movement

constructions and scrambling. In most cases, this is not because such constructions are

particularly amenable to a PF movement analysis but because they do not fit well i n the syntax.

This is not the case with PI. PI is a clearly defined movement operation, with a precise

phonological motivation and explicitly defined locali ty restrictions sensitive to phonological

information, which is generally not a characteristic of other putative examples of PF movement.

Strongest arguments for PI come from South Slavic cliti cization. However, a closer scrutiny

reveals that not only is PI not necessary to account for South Slavic cli ticization, but that South

Slavic cliti cization provides very strong evidence against it. I will focus here on SC; the reader is Ô Í Õ Í Ô Ô Í Ñ Ç È Ö È × Î È Ë Ð Ø Ù Ú Û Û Ü Ï Ý Õ È Ô Ñ Ð Þ ß à Þ Þ Ð È á È Õ â ã ä Ð Ç å Ô Í Þ æ Í ß Ç Ç È Ö à Ò garian and Macedonian, as

well as the li -construction across Slavic languages.

Notice first that we do not need PI to derive (16). SC is a language that freely allows left-ç Ô Ï á ß å Í è Ç Ô Ï ß Ç Ð È á Ù Þ Í Í Ö È × Î È Ë Ð Ø Ú Û Û Ü ß Õ È Ô Ñ Ð Þ ß à Þ Þ Ð È á È Õ Ò Í Õ Ç -branch extraction). Determiners,

and left-branches of NPs in general, can be routinely separated from nouns, as ill ustrated by (18),

which cannot be derived by PI and must involve syntactic movement (left-branch extraction) of

kojeg/tog. (More precisely, what (18), where the separation of kojeg/tog and é ê ë ì í î ï cannot be

accomplished by PI, shows is that we need to be able to separate determiners from nouns in SC

independently of PI.)

(18) Kojeg/Togi tvrdiš da su ti ð ñ ò ó ô õ ö ò ÷ ø ó ô ù ÷ ú which/that claim that are man seen

‘Which man do you claim that they saw.’

‘That man, you claim that they saw.’

û ü ý þ ÿ � � � � � ý � � � � � � ü � � � � ü � ÿ � � � ÿ þ � ü � ÿ � ü ü � � � � ü � ü � � � þ � � ü � � � þ � � � � � ÿ � � � � � � �with respect to cliti c placement in (15)-(16) is an artifact of the general possibili ty of left-branch

extraction in SC. According to them, in (16) we are also dealing with 1P placement, with the

phrase preceding the cliti c being located in front of the cliti c at SS after undergoing left-branch

extraction. What is important for our current purposes is that we do not need PI to derive (16),

which should be clear given the grammaticali ty of (18), which is underivable through PI.

Strong evidence against PI is provided by constructions in which a syntactically

immobile element attempts to host a cliti c. In (19) we have an element that apparently cannot

9

undergo syntactic movement. As shown in (20) and discussed by Progovac (1996) and Wilder � � � � � � � � � � ! � " # $ % & & ' & ( & � $ ) � * + & , $ ) - � � ' , - . � � � - $ / � & . & � & � , & . - � � / osition cliti c.5

(19) *Premai hodaju ti Mileni.

toward walk Milena.dat

‘They are walking toward Milena.’

(20) *Prema su Mileni hodali .

toward are Milena.dat walked

‘Toward Milena they walked.’

(21) cf. Oni su prema Mileni hodali .

they are toward Milena.dat walked

Given (21), it should be possible for the syntax to provide the following structure as input to PF:

(22) su [PP prema Mileni] hodali .

PI should then apply placing the cliti c after prema, incorrectly deriving (20).

Certain facts concerning split names discussed in Franks (1997) confirm that only

elements that can be placed in front of cliti cs by syntactic movement (or be base-generated in

front of cliti cs) can host cliti cs, which means that syntax, rather than phonology, provides a host

for SC cliti cs. In SC it is possible in some cases to inflect for structural case either one or both

names in first+last name complexes.6

� 0 1 " � 2 3 � � � 4 - ' , $ - 5 � 6 ) $ � ( 2

Leo.acc Tolstoi.acc read

‘Leo Tolstoi, I read.’ 7 2 8 3 � � � 4 - ' , $ - 5 6 ) $ � ( 2 Leo.acc Tolstoi.nom read

5Most SC prepositions are lexically unaccented (they procliti cize to the following stressed word)

and therefore cannot host cliti cs, which need a phonologically strong host. However, prema is

accented.

6Nominative is the default case in (23)-(25). Steven Franks (personal communication) suggests

treating the nominative elements in the constructions in question as frozen (i.e. uninflected)

forms.

10

9 : ; < = > ? @ A B ? C < D E B < F :

Leo.nom Tolstoi.acc read

The first name can be separated from the last name by movement only when they are both

inflected for structural case.

(24) a. Lava D E B < F > ? @ A B ? C < : G : H ; < = < D E B < F > ? @ A B ? C : 9 : H ; < = D E B < F > ? @ A B ? C < :

Significantly, cliti cization patterns with movement in the relevant respect.7

(25) a. Lava sam > ? @ A B ? C < D E B < @ a.

Leo.acc am Tolstoi.acc read

‘Leo Tolstoi, I read.’

b. *Lava sam > ? @ A B ? C D E B < @ < : Leo.acc am Tolstoi.nom read

c. *Lav sam > ? @ A B ? C < D E B < @ < : Leo.nom am Tolstoi.acc read

Franks observes that this is expected if only elements that can be base-generated or syntactically

moved in front of a cliti c can precede the cliti c. On the other hand, under the PI analysis we

would expect all of the constructions in (25) to be acceptable, since nothing blocks the derivation

in which the names remain in SpecIP overtly, with the cliti c located above the subject (C under

most PI analyses). PI would then place the cliti c after the first name, the first stressed word

following the cliti c, thus deriving (25), incorrectly predicting all of these constructions to be

good.

More evidence against PI and for strictly syntactic cliti c hosting is provided by

possibiliti es for contrastive focus in constructions involving 9 ? F I @ J K L < F J A M L ? B J N E L O ? P Q ? = E R(2001a). As shown in (26), either the first name or the last name of a complex town name split

by a cliti c can be contrastively focused.8 However, it is not possible to contrastively focus the

7Franks notes this with respect to examples in which only the first name is inflected for structural

case. As noted in Boškovi R (2000a), nothing changes if only the second name is inflected for

structural case.

8Under the most natural interpretation, one of the names split by a cliti c is contrastively focused.

11

whole complex name if the name is split by a cliti c. To do that, the cliti c has to follow the whole

name.9

(26) a. U GORNJI su Vakuf došli , ne DONJI.

in Gornji are Vakuf arrived not Donji

‘ In Gornji Vakuf they arrived, not Donji (Vakuf).’ S T U V W X Y Zsu TOPOLU došli , ne PALANKU. [ \ V W X Y W W ] ^ _ ` a ` b W W ] ] [ c ^ d e \ ` f g W b W \ Y W

h i \ V W X Y W _ ` a ` b W f j ^ k W ] ] [ c ^ d e \ ` f l V W Xka) Palanka.’

c. *U NOVI su SAD došli , ne ZRENJANIN.

in Novi are Sad arrived, not Zrenjanin

‘ In Novi Sad they arrived, not Zrenjanin.’

d. U NOVI SAD su došli , ne ZRENJANIN.

These facts are surprising if in split -name constructions the cliti c is placed after the first name by

PI, in which case (26) would have the following S-Structures.

(27) a. su u Gornji Vakuf došli ...

b. suZ V W X Y Z _ ` a ` b Z d ` m b [ T T T

c. su u Novi Sad došli ...

We could try to account for focus possibiliti es in these constructions by assuming that the

position immediately following the cliti c is a focus position, to which focused elements move.

The problem is to limit the focus requirement to the first name. It appears that nothing blocks the

derivation in which the whole complex town name is contrastively focused. By applying PI we

then incorrectly derive (26c), with the whole complex town name focused. (Note that the

This might be the reason why some speakers find constructions like (25a) somewhat degraded.

9Note that Donji Vakuf and n o p q o r o s o t q o exist. To the best of my knowledge, there is no Novi

Zrenjanin or Zrenjanin Sad. Contrastively focused elements are given in capitals. Note that non-

cliti c material can also intervene between the names, as in (i).

(i) U Gornji su oni Vakuf došli .

in Gornji are they Vakuf arrived

‘ In Gornji Vakuf, they arrived.’

12

preposition u is not stressed and therefore not a phonological word.)

(28) Syntax: su u NOVI SAD došli ...

PF: U NOVI su SAD došli ...

The data in (26) can be readily accounted for under purely syntactic movement accounts

of SC cliti cization. Apparently, there are two focus positions in SC, one above the auxili ary cliti c u v w x v y z y { x | } ~ � � y y � x � � x � } � � � � � z u v w � ~ � y � u v x � } � � � � � � � � � � � x � � � y � } � y { x � u ~ } x v � x � ~ � y � ypositions). The first option is ill ustrated by (26a,d) and the second by (26b).10 Notice that if a

cliti c host can be placed in front of the cliti c only through syntactic movement (I am ignoring

here elements that are base-generated in front of cliti cs), we have a direct correspondence

between PF word order and the output of the syntax. Since neither the pre-auxili ary nor the post-

auxili ary focus position contains the whole complex name in (26a-c), the whole complex name

cannot be focused, only its parts can be. In (26d), the whole name can be focused, since the

whole name can be located in the focus position in the syntax.

I conclude, therefore, that split -name constructions, which have previously been argued

to provide evidence that cliti c placement cannot be syntactic (see, for example, Anderson 1996),

are not only consistent with syntactic placement accounts, but in fact provide strong evidence

against PI.11

To summarize the discussion in this section, the data discussed above indicate that only

elements that can be independently shown to be able to undergo syntactic movement can precede

and host cliti cs in SC. The Prosodic Inversion analysis should be rejected since it fails to capture

the correlation between syntactic movabili ty and the abili ty to host a cliti c. I conclude, therefore,

that the mechanism of Prosodic Inversion is not available in natural language. Since Prosodic

Inversion was probably the strongest argument for PF movement, it is tempting to conclude that

there is no PF movement at all (for mu� � � y { y � u v ~ w } � � � � � } x v � � y y � x � � x � } � � � � � u � �

10I assume that it is not possible to “activate” both focus positions in the same clause. Doing this

would often lead to a relativized-minimali ty violation with focus movement (i.e. focus

movement across a focused element). Notice also that constructions involving multiple clause-

mate foci are extremely rare crosslinguistically, most languages disallowing them (see Kiss

1995).

11The same holds for its non-movement versions that place cliti cs after the first word post-

syntactically (see Anderson 1993 and Caink 1998).

13

1.3 Clitic placement

In this section I address the issue of cliti c placement. Until recently, it was standardly assumed � � � � � � � � � � � � � �   ¡ � � ¢ � £ � � ¤ � � � ¡ ¥ � ¦ � § ¨ �   © ª ¡ ¥ � © � � ¦ � � � � � � « ¤ � � � � ¤ ¦ � ¦ ¬ ¤ ­ ® ¤ ¯ � ° ± ² ³95). Since

then, a number of arguments against it have been given in the literature (see, for example, ¬ ¤ ­ ® ¤ ¯ � ° ´ µ µ µ � ¶ ´ µ µ ² � ¶ � � � ¦ ® ² ³ ³ · ¶ ¸   ¤ © ¤ ¯ � � ² ³ ³ ³ ¶ � � ¹ ¡ « � ¦ ¤ ¯ � ° ² ³ ³ · � ¶ º ¶ ² ³ ³ ³ ¶ � ¦ ¥ » � � ¥ ¡   � ¦ ¥¼ � ¯ �   ² ³ ³ ½ ¾ § � � ¦ � ¡ � � ¡ � � � � � � �-in-C position seems to be completely abandoned, I will not dwell

on it here.

I will focus on an analysis which shares with the cliti cs-in-C analysis the assumption that

SC cliti cs are located in the same head position. The analysis in question, which was proposed

independently in Franks (1998) and Caink (1998) (see also Franks 2000a, Franks and King 2000,

and Caink 1999), holds that SC clitics are located as high as possible, i.e. in the highest head « ¤ � � � � ¤ ¦ ¤ ¿ � � ¡ �   � � � ª � ¡ § À ¤ © ¡ � � ¡   ¢ � � � � ¤ Á ¡   ¡ � ¡ ¦ � ¢ ¤   ® ± � ¡ ¡ ¬ ¤ ­ ® ¤ ¯ � ° ² ³ ³ ½ � � ¦ ¥   ¡ ¿ ¡rences

therein), the authors in question assume that only phrase structure which is independently

required is projected. As a result, since clauses do not always have the same phrase structure

projected, under their analysis SC cliti cs do not have a fixed structural position. They can also

end up being pretty low in the structure (by low I mean lower than CP, were it projected). As a

result, several problems that arise under the cliti cs-in-C analysis do not arise under the cliti cs-in-

the-highest-head-position-of-the-clause analysis. However, this analysis has some problems of

its own. Focusing on Franks’s version of the analysis on which cliti cs undergo overt movement � ¤ � � ¡ � � © � ¡ � � � ¡ � ¥ « ¤ � � � � ¤ ¦ ¤ ¿ � � ¡ �   � � � ª � ¡ ¶ � � ¥ � � � ª � � ¡ ¥ � ¦ ¬ ¤ ­ ® ¤ ¯ � ° ± ´ µ µ ² � ¾ ¶ � � � � ¯ ¡   £ difficult

to implement this analysis. In particular, there is no principled way in the current theory to

ensure that SC cliti cs always move overtly to the highest head position projected. Franks

assumes that the movement is driven by a strong feature of cliti cs. Since cliti cs do not have a

fixed structural position, it must then be the case that the strong feature is checked through

movement to different positions in different clauses. It is very difficult to see how this state of

affairs can be formalized� ¦ � «   � ¦ � � « � ¡ ¥ ¢ � £ ± � ¡ ¡ ¬ ¤ ­ ® ¤ ¯ � ° ´ µ µ ² � ¿ ¤   � ¥ ¥ � � � ¤ ¦ � � «   ¤ º � ¡ Á �

concerning the successive cyclic nature of the cliti c movement and the relation between cliti c-

and V-movement). However, the most serious problem with this analysis is that, just like the

cliti cs-in-C analysis, the analysis in question crucially relies on the assumption that clause-mate

cliti cs are all l ocated in the same head position. There is considerable amount of evidence that

the assumption is untenable, which means that any analysis that cruciall y relies on it must be

rejected on empirical grounds. I discuss the relevant evidence in the next section. The discussion

will also give us an insight into the actual structural positions of SC cliti cs.

14

1.3.1 SC clitics do not cluster in the same head position

Â Ã Ä Å Æ Ç È É Ê Ë Ì Í Î Ï Ï Ð ÇÑ Ò Ó

Æ Ô É Ê Ë Õ Å Ö Å Ê Ë Õ Å È × Å Ã Ø Ç Ã × Ù Ç Ú Ö Å-mate cliti cs in SC do not cluster in the same

head position in the syntax based on VP elli psis. She observes that VP elli psis in SC can delete

part of a cliti c cluster, leaving some cliti cs behind.12

(29) a. Mi smo mu ga dali , a i vi ste mu ga dali (takodje).

we are him.dat it.acc given and also you are him.dat it.acc given too

‘We gave it to him, and you did too.’

b. cf. *Mi smo mu ga dali , a i vi mu ga ste dali (takodje).

The possibili ty of VP elli psis in (29a) raises a serious problem for the assumption that cliti cs

cluster under the same node in SC. Under this assumption, (29a) has to involve deletion of a non-× É È Ö Ã Ë Ã Ú Å È ÃÑ Û

Ø Ë × Ø Ë Ö Ö Ã Ç È Õ Ç Ô Õ Ù Ü Ç Ö Ö Ú Ý Å Õ È É Ã Ã ÉÒ

Å Æ É Ö Ö ËÒ

Ù ÅÞ

Â Ã Ä Å Æ Ç È É Ê Ë Ì Ã Ç ß Å Ö Í à Ï ÇÓ

Ç Ö Å Ê Ë Õ Å È × Åthat SC cliti cs are not located under the same node in the syntax, i.e., they are located in separate

maximal projections, the auxili ary cliti c being higher than the pronominal cliti cs. (The

ungrammaticali ty of (29b) is consistent with this assumption.) Example (29a) can then be

analyzed as involving constituent deletion.

Another argument that clause-mate cliti cs in SC are not located in the same head position Ë Ö Æ Ô É Ê Ë Õ Å ÕÒ

Ü á Ë Ù Õ Å Ô Ç È Õ â Ç Ê Ç Ô Í Î Ï Ï ãÓ Ò

Ç Ö Å Õ É È × Ù Ë Ã Ë × Æ Ù Ç × Å Ý Å È Ã Ë È × É É Ô Õ Ë È Ç Ã Ë É È Ö Ù Ë ß Å(30).

(30) Ivan je kupio auto i razbio ga.

Ivan is bought car and ruined it

‘ Ivan bought a car and ruined it.’

If clause-mate cliti cs must be located in the same head position, (30) must involve coordination

of two clauses. This means that the second conjunct must contain a deleted auxili ary, as shown in

the structure

Ë È Í ä ÎÓ Þ

Í á Ë Ù Õ Å Ô Ç È Õ â Ç Ê Ç Ô Î Ï Ï ã Ç Ô Å å É Ù Ù ÉÛ

Ë È æ Ã Ø Å Ö Ã Ç È Õ Ç Ô Õ Ç Ö Ö Ú Ý Æ Ã Ë É È É å Ö Ü È Ã Ç × Ã Ë ×accounts of that time that cliti cs are located in C. Nothing, however, changes if (30) involves

clausal coordination on the IP, rather than the CP level. Notice also that by itself, Razbio ga je is

acceptable.)

12The constituent undergoing elli psis in (29) could actually be larger than VP. (The same remark

applies to the VP coordination and the VP fronting data discussed below.)

15

(31) [CP Ivan je kupio auto] i [CP [C0 razbio ga je] [ IP pro...]]

On the other hand, if clause-mate cliti cs in SC do not have to be located in the same head

position (more precisely, if auxili ary cliti cs can be higher in the structure than pronominal

cliti cs), (30) can be given the structure in (32), with no deleted auxiliary in the second conjunct ç è é é ê ë ì í ë î ï ð ñ ò ò ó ô õ ë ö ô ÷ ë ö é ø é ù ô ï ú é ø ø ï è û ü è è ï ë ý ë õ ù þ é è ù ö ü û ù ü ö é ë õ ç ÿ ó � � � 13

(32) [CP Ivan je [VP kupio auto] i [VP razbio ga]]

� ï ú ø é ö ô ý ø � ô î ô ö ë � è é ö î é ù þ ô ù ù þ é ø é ú é ù ï ë ý ë õ ù þ é ô ü � ï ú ï ô ö � ï ý ç ÿ ó � è þ ë ü ú ø ý ë ù � é ô ú ú ë � é ø � � ùviolates the condition on elli psis in (33), whose effect is ill ustrated by the impossibili ty of

deleting the cliti c in the second conjunct of Spanish (34a).

(33) No part of an X0 may be deleted (forward deletion)

(34) a. *Juan lo compró y Javier lo leyó.

Juan it bought and Javier it read

‘Juan bought it and Javier read it.’

b. cf. Juan lo compró y Javier lo leyó.

It also violates Wilder’s (1997) condition on forward deletion given in (35), since the auxili ary is

preceded and presumably c-commanded (given Kayne’s 1994 LCA) by the participle. The effect 13An anonymous reviewer proposes another analysis on which the pronominal element moves out

of the second conjunct, with ga in the second conjunct treated as a resumptive pronoun which

saves the construction from an island violation. There are several problems with this analysis.

Thus, one question that arises is why the pronoun is also not pronounced in the raised position

outside of the coordination given that the resumptive pronoun strategy normally does not prevent

the head of the relevant chain to be phonologically realized. The most serious problem, however,

concerns the fact that SC does not have the resumptives-as-island-rescuers strategy, as ill ustrated

by (i) for the island under consideration.

(i) *Auto, Jovan jeí ü ï ë í ü ð ü ï ö ô � ï ë

ga.

car Jovan is bought house and ruined it

‘Jovan bought a house and ruined a car.’

16

of the condition is ill ustrated by (36). (Examples (36b-c) are from SC. Notice that a parallelism

constraint on deletion requires that the deleted element’s position is the same as the

antecedent’s.) For more empirical evidence for the condition, see Wilder (1997) and Wilder and � � � � � � � � � � �14

(35) Head Condition: no constituent can be deleted that is c-commanded by an overt X0 in its

conjunct at S-Structure.

(36) a. *John has bought Mary a book and given Mary a book.

b. * Ivan kupuje Mariji knjige i daje Mariji knjige.

Ivan buys Marija books and gives Marija books

c. cf. Ivan Mariji knjige kupuje i Mariji knjige daje.

Wi � � � � � � � � � � � � � � � � � � � ! � ! � � � � � � " � � � � � # � � � � � � � $ � � % � � � � ! � & ' ( � � � � $ � � $ � �analysis of (30), shown in (32), since if the second conjunct is a VP (or AgroP for that matter), it

does not have to contain a deleted auxili ary. Notice, however, that the clause-mate auxili ary and

the pronominal cliti cs are not located in the same head position in (32), the auxili ary cliti c being

higher than the pronominal cliti c.

) � * � + , � � $ - . / 0 0 0 � 1 / 0 0 2 � 3 ) " � � � $ � � � � $ � � � ( � ! � 4 5 ( � $ $ ( � � � � � � � � ( � � � $ � ! � � � # �head position based on constructions like (37), where the presence of a parenthetical makes a

cliti c split possible:

(37) a. Oni su, kao što sam vam rekla, predstavili se Petru.

they are as am you.dat said introduced self.acc Petar.dat

‘They, as I told you, introduced themselves to Petar.’

b. *Oni se, kao što sam vam rekla, predstavili su Petru.

As a result of the presence of the parenthetical, which introduces an additional I-phrase (an I-

phrase boundary immediately follows the parenthetical), each cliti c in (37a) is located in the

second position of its I-phrase, satisfying (5).15 What is important for our current purposes is that

14As discussed in Boškovi - (2001a), I assume that (33) and (35) apply to elli psis deletion, but

not to copy deletion, discussed below.

15The example would be unacceptable without the parenthetical because the second cliti c would

not be located in the second position of its I-phrase. The unacceptabili ty of *Oni su predstavil i se

Petru and the acceptabili ty of (37a) show that I-phrase-mate, but not clause-mate cliti cs, have to

cluster together, which indicates that the cliti c clustering requirement is prosodic in nature, not

17

clause-mate cliti cs in (37a), su and se, are not located in the same head position. The auxili ary

cliti c is clearly located higher in the structure than the reflexive cliti c. The acceptabili ty of (37a)

provides strong evidence against the assumption that SC cli tics cluster together in the same

position in the syntax.

More evidence against this assumption is provided by VP-fronting data noticed by Damir 6 7 8 7 9 : ; < 9 = > ? 7 @ A > B B C ? D A 7 E D > ? F 7 ? G H D @ G < 9 7 ? G 6 7 8 7 9 : I J J K F L M N < O > P = < 9 8 < E N 7 E = ; < 7 Q < 9 = R N >allow VP fronting with auxili ary cliti cs also allow VP fronting to split the cliti c cluster, which

clearly shows that clause-mate cliti cs do not have to cluster together.16

(38) Dali ga Mariji su Ivan i Stipe.

given it.acc Marija.dat are Ivan and Stipe

‘Give it to Marija, Ivan and Stipe did.’

Example (38) provides further evidence that auxiliary cliti cs are higher in the structure than

pronominal cliti cs.

A different type of argument that SC clitics are not located in the same position in the

syntax is provided by subject-> 9 D < ? E < G = < ? E < ? E D 7 @ 7 G 8 < 9 P = L S = G D = A C = = < G D ? T > U Q > 8 D V : W X X I 7 F Y

auxili ary cliti cs can be higher than such adverbs, as indicated by the availabili ty of the sentential-

subject reading of the adverb in (39). Significantly, pronominal object cliti cs cannot occur above

subject-oriented adverbs. Example (40) is fully acceptable only on the manner reading.

(39) Oni su pravilno odgovorili Mileni.

they are correctly answered Milena.dat

‘They did the right thing in answering Milena.’

‘They gave Milena a correct answer.’

syntactic (i.e., it does not follow from a requirement that clause-mate cliti cs be located in the

same position syntacticall y). See in this respect BoškoviV (2000a, 2001a) and the discussion in

section 1.1. Notice that Oni su se, kao što sam vam rekla, predstavili Petru is also acceptable.

16Some speakers do not allow fronting of the complement of an auxili ary cliti c at all . For relevant

discussion, see BoškoviV (2001a), Browne (1975), Caink (1998), Schütze (1994), Tomi

V (1996), 7 ? G H D @ G < 9 7 ? G 6 7 8 7 9 : I J J Z 7 F L [ E 7 ; ; < 7 9 = E N 7 E \ 9 > 7 E D 7 ? Y P C E ? > E ] < 9 P D 7 ? Y 7 @ @ > R = D E L S = ? > E < G D ? ^ ? L

35, speakers who accept (38) also accept Dali Mariji su ga Ivan i Stipe. See section 1.4. for

relevant discussion.

18

(40) Oni su joj pravilno odgovorili .

they are her.dat correctly answered

‘ *They did the right thing in answering her.’

‘They gave her a correct answer.’

Apparently, auxili ary clitics can occur higher than subject-oriented adverbs. Pronominal object

cliti cs, on the other hand, cannot. It must then be the case that the two do not occur in the same

structural position. More precisely, auxili ary cliti cs must be higher in the structure than

pronominal cliti cs.17

There is also evidence that pronominal cliti cs themselves are not located in the same head _ ` a b c b ` d e f g h a i j c k l _ m d ` n b o p q r r s m i t i q r r r u ` t a l v n l a c g m c i m a b w w h a c v m c l x b d p y q u i z ` d a c v h z c b ` d a b dwhich VP elli psis leaves behind a dative cliti c while eliding an accusative cliti c are acceptable,

whereas constructions in which VP elli psis strands an accusative cliti c and elides a dative cliti c

are unacceptable.

(41) a. ?Mi smo mu ga dali , a i vi ste mu ga dali (takodje).

we are him.dat it.acc given and also you are him.dat it.acc given too

‘We gave it to him, and you did too.’

b. *Mi smo mu ga dali , a i vi ste ga mu dali (takodje).

j c k l _ m d ` n b o b d c l v _ v l c a c g l z ` d c v m a c m a b d x b z m c b d { that the dative and the accusative cliti c are

located in different maximal projections, the dative cliti c being structurally higher than the

accusative cliti c.

More evidence for the conclusion is provided by the cliti c climbing data discussed in

Stjepan n b o p q r r s t i q r r r u e | v ` { ` n m z p q r r } u a g ` ~ a c g m c z w b c b z z w b � t b d { b a � m v { b d m w w � _ ` a a b t w l ` h cof the finite complement embedded under verbs like � � � � � � � ‘want’ , as shown in (42) � � � � � � � � � � � �observes that if the da-clause embedded under � � � � � � � contains two pronominal cliti cs, it is

possible to climb only one of the cliti cs into the matrix clause. When this happens in

constructions containing a dative and an accusative cliti c, the dative cliti c is the one that moves

into the matrix clause (see (43)).

17One could try to use sentential adverbs on a par with parentheticals to split a cli tic cluster.

However, an interfering factor here is that, in contrast to parentheticals, sentential adverbs like

pravilno are not naturally parsed as separate I-phrases, which is a prerequisite for using them to

split a cliti c cluster.

19

(42) a. M � � � � � � � � � � ga vidi.

Milan wants that him sees

‘Milan wants to see him.’

b. ?Milan ga � � � � � � � � � � � (43) a. ?Marija mu � � � � � � ga predstavi.

Marija him.dat wants that him.acc introduces

‘Marija wants to introduce him to him.’

b. *Marija ga � � � � � � mu predstavi.

� � �� � � �   � � ¡   ¢ £ � ¤ � � £

� ¥�

� � ¥� ¦   �

�¤ � £

�� � § ¨ © ª ¤ � � � � � « ¬   � �   ­ £ � ¬

� ¥� � �

�� � � ¦ � �

�� ¦ � £ £

�¤ ® ¦

�® ¤ � � � «

higher than the accusative cliti c. Example (43b), where the accusative cliti c skips the dative

cliti c, then involves a familiar relativized minimali ty violation.

¯¥

� ¬   � �   ­ � � ° ¦   ��

¤ � £�

¬ ¤   ± ²   ³ ´   � � ¡ § µ ¶ ¶ · � ª ¦   � ¬ � ¤ ± £� ¥

�� � ¥

� � ��

� � � ¦ � ��

� ¦ � £¥

� °¥

� ¤than the accusative cliti c:

(44) ??Oni su mu, kao što sam vam rekla, predstavili ga

�® ¸ � �

they are him.dat as am you.dat said introduced him.acc yesterday

‘They, as I told you, introduced him to him yesterday.’

(45) *Oni su ga, kao što sam vam rekla, predstavili mu

�® ¸ � �

The above data provide evidence that the auxili ary cliti c, the dative cliti c, and the

accusative cliti c are located in different positions, the auxili ary cliti c being higher than the

pronominal cliti cs and the dative cliti c being higher than the accusative cliti c. Given that, as

£ ® ° ° � £�

� � � � ¹ ��

� � � ¢ � § · º º © ª � � � ²   ³ ´   � � ¡ § · º º » ¼ · º º ½ � ª ¼ £ ® ¢�

� ¦�-oriented adverbs are TP-

adjoined,18 the data in (39)-(40) provide evidence that the auxili ary cliti c is located above TP,

and the pronominal cliti cs below TP. All the data considered above can be straightforwardly

accounted for if auxili ary cliti cs move (or can move) overtly to Agrs, which is higher than TP,

and pronominal object cliti cs are located in their Case-checking agreement projections (AgroPs),

which are lower than TP. The fact that the dative cliti c is higher than the accusative cliti c can be

readily captured if the dative cliti c is located in AgrioP and the accusative cliti c in AgrdoP, ¾ ° ¤ �   ¿ ¢ � � � °¥

� °¥

� ¤ � �� ¥

� £�

¤ ® ¦�

® ¤ �� ¥

� � ¾ ° ¤ �   ¿ � §�

� � � � £   ²   ³ ´   � � ¡ µ ¶ ¶ ¶ � ¼ µ ¶ ¶ · � � � �� � �� � � �   � � ¡ · º º À � ¼ ¢ ¼ · º º º � ¾ £ � ± � � � ¤ � � � � « £ � £ ­ � £ � ¤   �   £ � � ¬   ¤ Á  � ¦

¥¢ « ¯   ± � � · º º º � ª 19

18See Cinque (1998) for a much more detailed discussion of position of adverbs.

19Another possibili ty in the multiple-specifiers framework is that the dative and the accusative

cliti c are located in distinct specifiers of the same head, possibly Agro or Chomsky’s (1995) v,

20

(46) [AgrsP auxili ary-cliti c [AgrioP dative cliti ci [AgdroP accusative cliti cj [VP ti main verb tj ]]] ]

The current analysis provides a principled explanation of Browne’s (1975) observation

that second position cliti cs cannot occur as complements of a preposition in SC, ill ustrated by the

ungrammaticali ty of (47).20

(47) *Prema mu Ã Ä Å Æ Ç

toward him.dat run

‘They are running toward him.’

È É Ê Ë É Ì Í É É Æ Ê Ë Î È Ï Æ Ð É Ñ Ò Ó Ó Ô Õ Ö Õ Î Ê × Ø Ù Ú Ø Û Ë Ü Ñ Ò Ó Ó Ô Õ Ö Ý Ã Þ Æ Í Î ß Ä Õ à à Õ Ã Ë Ì Õ Ð Ë Ã á Ø â Ñ ã ä Ö Ì Ø Î â Ë Ä à É Ã Þ Æclaim that SC pronominal cliti cs must move to their Case-checking position overtly. Suppose

that, as argued in Watanabe (1993), AbeÐ É Ñ Ò Ó Ó Ô Õ Ö Ý Õ Î Ê × Ø Ù Ú Ø Û Ë Ü Ñ Ò Ó Ó Ô Õ Ý Ò Ó Ó Ò Ì Ö Ý å Õ É Æ -checking

within a traditional PP takes place in an AgrP dominating the PP. (This is on a par with Case-

checking “within” VP and TP.) The problem with (47) is then that the cliti c did not move to its

Case-checking position overtly. (The problem does not arise in (21), where I assume Mileni does

not have to move to its Case-checking position overtly.) Notice that moving the cliti c in (47) to

with the dative cliti c being located in the higher and the accusative cliti c in the lower specifier.

Notice that although the analysis in the text assumes Agr phrases, it can be easily restated in an

Agr-less framework. Notice also that main verbs in SC can undergo short V-movement, which I

ignore here (see BoškoviÜ 2001a and Stjepanovi

Ü 1998c, 1999 for relevant discussion).

20Browne (1975) observes that SC has a set of non-second position accusative cliti cs which can

occur as complements of prepositions taking accusative complements. (This usage is archaic.)

(i) Marko gleda na nj.

Marko looks on him

‘Marko is looking at him.’

The claim in the text concerning location in Case-checking positions in overt syntax does not

refer to these elements. It refers only to second position cli tics. (Browne actually does not call

the pronominal element in (i) a cliti c. He observes an interesting fact that na nj is stressed,

though the preposition na on its own is a proclit ic. He also observes that the element in question

can be conjoined.) The reader is also referred to Franks (2000b) for discussion of cliti cs as

complements of a preposition in the context of cli tics within NP.

21

SpecAgrpP overtly, as in (48), does not help since preposition stranding is not possible in SC, as

shown in (49).21

(48) *da [AgrpP mui [VP prema ti æ æ ç è é e.

that him.dat toward run

(49) a. *Komei ê ë ì ç è é e [PP prema ti]?

who.dat they run toward

‘Who are they running toward?’

b. cf. Prema komei oni tr é e ti?

Having established the structural position of auxili ary and argumental pronominal cliti cs,

I turn now to the ethical dative cliti c. As discussed in Boškovi í (2001a), in contrast to

argumental pronominal cliti cs (see (51)), ethical dative cliti cs (see (50)) can occur above subject-

oriented adverbs. Thus, in contrast to (51), in (50) the adverb can have the sentential as well as

the manner reading. (It is diff icult to translate ethical dative into English so I ignore it in the

translations. See below for discussion of its semantics.)

(50) Oni su ti pravilno odgovorili Mileni.

they are you.dat correctly answered Milena.dat (you=ethical dative)

‘They did the right thing in answering Milena.’

‘They gave Milena a correct answer.’

(51) Oni su joj pravilno odgovorili .

they are her.dat correctly answered

‘ *They did the right thing in answering her.’

‘They gave her a correct answer.’

Ethical dative then must be structurally higher than argumental dative and accusative cliti cs. This

is not surprising. While argumental pronominal cliti cs are closely related to the verb, ethical î ï ç ì ð ñ ò ó ì ç ì ò ô ï è ñ ë ê ç õ ö ï î ï ë ê ð ì í -÷ ê ò ì í ø ù ú û û ü ì ë ý ï ò ç ô þ ÿ ÿ ñ ô ç ô ç � ï ç ç � ñ ñ ç � ì ò ï ó dative is a

sentential particle. It is then no surprise that it is structurally higher than argumental pronominal

ò ó ì ç ì ò ô õ ø � ô î ì ô ò þ ô ô ñ î � � ö ï î ï ë ê ð ì í -÷ ê ò ì í � ç � ñ ñ ç � ì ò ï ó î ï ç ì ð ñ � ï ô ï ë ñ ë î ñ ï è ì ë ÿ � þ ï ó ì ç � õ � ç ô 21The exact nature of the ban on preposition stranding in overt syntax in SC does not affect the

point made here. For much relevant discussion, see Abels (2001a,b) and Boškovi í (2002c).

22

pragmatic function is to express closeness and sympathy between the speaker and the addressee,

or to incite the hearer’s attention and involve him or her in the narration.)

Notice also that as expected given the adverb data presented above, when both an ethical

and an argumental dative are present in a sentence, the ethical dative must precede the

argumental dative. This is ill ustrated in (52).22

� � �� � � e sam ti joj pomogla.

yesterday am you.dat her.dat help (you=ethical dative)

‘Yesterday, I helped her.’

b. *Ju� e sam joj ti pomogla.

I conclude therefore that ethical dative is located higher in the structure than argumental

pronominal cliti cs. Given that ethical dative stil l follows auxiliary cliti cs, which are suggested

above to be located in Agrs, and that it is higher than sentential adverbs, which are argued to be

TP-adjoined in Watanabe (1993) and Boškovi � (1997a), I tentatively propose that the ethical

dative cliti c is located in a discourse- � � � � � � � � � � � � � � � � � � � � � � � � � ��

� �� ! " # $ % & '

reader should

not attach too much importance to the label), located above TP but below AgrsP. The following

structure then gives us the basic position of the elements discussed above.23

(52) [AgrsP auxili ary cli tic [ ( P ethical dative [TP sentential adverb [AgrioP dative cliti ci [AgrdoP

accusative cliti cj [VP ti main verb tj ]]]] ]]

The data discussed in this section strongly argue against morphological template analyses

of the order of cliti cs within the cliti c cluster (see Halpern 1995 and Schütze 1994, among

others).24 This type of analysis views the cliti c cluster as a linearly ordered set of optional slots

into which morphemes bearing certain feature combinations are placed. Under the morphological 22See Franks and King (2000), Fried (1999), and Toman (1999) for the corresponding data from

Czech. Notice that (52b) is ungrammatical because the first dative cannot be ethical dative,

ethical dative being limited to the 1st and 2nd person pronouns.

23In section 1.4. I address the issue of where the auxili ary cliti c is base-generated.

24The same holds for analyses in which the cliti c cluster is ordered through arbitrary optimali ty-

theoretic constraint rankings (see Anderson 1996, who outlines such an analysis, and Legendre

1999, 2000, who fleshes it out with respect to Bulgarian and Macedonian cliti cs, which are

discussed in section 2).

23

template view, the ordering of cliti cs within the cliti c cluster is essentially arbitrary; it does not

follow from anything. The syntactic account of the order of cliti cs is more principled. Under this

account the order of cli tics within the cluster matches the structural height of the cliti cs. The

above facts strongly indicate that this is indeed the case.25 The syntactic account of the ordering

of cliti cs within the cliti c cluster is obviously conceptually preferable to an arbitrary

morphological template which would mirror the syntax by accident. Under the syntactic analysis,

the order of cliti cs within the cluster ultimately follows from the hierarchical arrangement of

projections where they are located, which seems to be universal. Thus, under the syntactic

account, the order dative cliti c-accusative cliti c follows from the plausibly universal AgrioP-

over-AgrdoP hierarchy, which is moreover not cliti c specific (see, for example, Lasnik 1995).

Under the morphological template approach, we need language specific and/or cliti c specific

mechanisms which mirror syntax by accident to get the order dative cliti c-accusative clit ic.

The morphological template analysis was originall y proposed to handle idiosyncrasies of

cliti c ordering that seem problematic for the syntactic view. (Notice that the morphological

template analysis does not explain the idiosyncrasies; it merely provides a formal way of stating

them.) The major idiosyncrasy of SC cliti c ordering concerns the third person singular auxili ary

cliti c je ‘ is’ , which, in contrast to other auxili ary cliti cs, follows pronominal cliti cs within the

cliti c cluster. In section 1.4 I will show that this behavior of je is not an accident and that it is

consistent with the syntactic view of cliti c ordering.

In light of the facts discussed in this section, I conclude that SC cliti cs are located in

different projections in the syntax. The order of cliti cs within the cluster matches their structural

height, which calls for a structural account of the order. Pronominal cliti cs in SC are located in

their Case-checking positions overtly.

1.4 Je

We have seen in section 1.3. that auxili ary cliti cs precede pronominal cliti cs. They are also

higher in the structure than pronominal cliti cs. The third person singular auxiliary cliti c je differs

from other auxili ary cliti cs in that it must follow pronominal cliti cs, as ill ustrated in (54).

(54) a. Oni su mu ga predstavili .

they are him.dat him.acc introduced

25Boškovi ) (2001a) and Franks (1998) show that more subtle cliti c orderings that were not

discussed in this work are also a result of a hierarchical arrangement of functional projections

that house the cliti cs in question.

24

‘They introduced him to him.’

b. Ona mu ga je predstavila.

she him.dat him.acc is introduced

c. *Oni mu ga su predstavili .

d. *Ona je mu ga predstavila.

As noted above, the idiosyncratic behavior of je is often cited as an argument for arbitrary

morphological template analyses of the order of cliti cs within the cliti c cluster, which were

argued against in the previous section. In this section I will provide a principled, structural

account of the idiosyncratic behavior of je.

Interestingly, as shown in Boškovi * (2001a), when applied to je, the tests run in section

1.3. with respect to other auxili ary cliti cs show that in the syntax, je is higher than pronominal

cliti cs, just like other auxiliary cliti cs. More precisely, evidence from VP elli psis, VP fronting,

parenthetical placement, and subject-oriented adverbs placement strongly indicates that in the

syntax, je is higher than pronominal cliti cs (see (55)-(58)), just like other auxili ary cliti cs (see

(59)-(62)). (The reader is referred to section 1.3.1 for a more detailed discussion of the tests in

question.)

(55) a. Ona mu ga je predstavila, a i on je mu ga predstavio.

she him.dat him.acc is introduced and also he is him.dat him.acc introduced

‘She introduced him to him and he did too.’

b. *Ona mu ga je predstavila, a i on mu ga je predstavio.

(56) a. Dao ga Mariji je Ivan.

given it.acc Marija.dat is Ivan

‘Give it to Marija, Ivan did.’

b. *Dao je Mariji ga Ivan.

(57) a. ?#On je, #kao što sam vam rekla#, predstavio ga Petru#.

he is as am you.dat said introduced him.acc Petar.dat

‘He, as I told you, introduced him to Petar.’

b. *#On ga, #kao što sam vam rekla#, predstavio je Petru#.

(58) a. Jovan je pravilno odgovorio Mileni.

Jovan is correctly answered Milena.dat

‘Jovan did the right thing in answering Milena.’

‘Jovan gave Milena a correct answer.’

b. On joj je pravilno odgovorio.

he her.dat is correctly answered

25

‘ *He did the right thing in answering her.’

‘He gave her a correct answer.’

(59) a. Vi ste mu ga predstavili , a i mi smo mu ga predstavili .

you are him.dat him.acc introduced, and also we are him.dat him.acc introduced

b. *Vi ste mu ga predstavili , a i mi mu ga smo predstavili .

(60) a. Dali ga Mariji su Ivan i Stipe.

given it.acc Marija.dat are Ivan and Stipe

‘Give it to Marija, Ivan and Stipe did.’

b. *Dali su Mariji ga Ivan i Stipe.

(61) a. ?#Oni su, #kao što sam vam rekla#, predstavili ga Petru#.

they are as am you.dat said introduced him.acc Petar.dat

‘They, as I told you, introduced him to Petar.’

b. *#Oni ga, #kao što sam vam rekla#, predstavili su Petru#.

(62) Oni su pravilno odgovorili Mariji .

they are correctly answered Marija.dat

‘They did the right thing in answering Marija.’

‘They gave Marija a correct answer.’

Examples (55a) and (56a) show that VP ell ipsis and VP preposing can affect pronominal cliti cs

without affecting je. Affecting je by these processes without affecting pronominal cliti cs leads to

ungrammaticali ty, as ill ustrated by (55b) and (56b). The VP elli psis and VP fronting data show

that je is higher than pronominal cliti cs in the syntax. The fact that, as il lustrated by (58), je can

occur above subject-oriented adverbs (when it does not follow a pronominal cliti c) while

pronominal cliti cs cannot points to the same conclusion. That je is indeed higher than

pronominal cliti cs in the syntax is conclusively confirmed by the contrast between (57a) and

(57b). In all these respects, je behaves like other auxili ary cliti cs (see (59)-(62)). Recall ,

however, that, as shown in (54), in contrast to other auxili ary clitics, in the final PF

representation je follows pronominal cliti cs in the cliti c cluster.

This state of affairs is surprising. We have seen that the relative structural height of cliti cs

matches the order of cliti cs within the cliti c cluster, the highest cliti c in the cliti c cluster being

first in the linear order. This state of affairs can be easily accounted for under the structural

account of cliti c ordering. From this perspective, je exhibits a schizophrenic behavior: it is higher

than pronominal cliti cs in the syntax, but it follows them (which means is lower than them) in the

phonology. How can the schizophrenic behavior of je, noted in Boškovi + (2001a), be accounted

for? Previous accounts of the cliti c order in (54) cannot account for the data in (55)-(58). Tomi +

26

(1996), Franks and King (2000:329-330), and Franks and Progovac (1994) account for (54) by

placing je and other auxili ary cliti cs in different positions syntactically; in particular, by placing

je below pronominal cliti cs and other auxili ary cliti cs above pronominal cliti cs at S-Structure.26

The data in (55)-(58) show that je is higher than pronominal cliti cs in the syntax.

So, how can we account for the dual behavior of je with respect to syntax and

phonology? Apparently, somewhere in PF je and adjacent pronominal cliti cs are somehow

“switched”. Given all the arguments against PI in section 1.2. and more generally arguments

against PF movement in Boškovi , (2001a), it would be desirable to achieve “ the switch” without

actual PF movement. It is shown in Boškovi , (2001a) that the dual behavior of je can be

accounted for in a principled way without any PF movement under Franks’s (1998, 2000a)

approach to the pronunciation of non-trivial chains.

It is standardly assumed that only the head of a non-trivial chain can be pronounced.

However, Franks (1998) (for relevant discussion and examples, see also Bobalji k 1995,

Boškovi , 2000b, 2001a, 2002b, Boškovi , and Franks 2002, Franks 2000a, Hiramatsu 2000a,b,

Lambova 2002, and Pesetsky 1997, 1998) proposes that a lower copy of a non-trivial chain is

pronounced in PF iff this is necessary to avoid a PF violation.27 One argument for the proposal

given in Boškovi , (2000b, 2002b) involves multiple wh-fronting.

A number of languages require all wh-phrases to be fronted in questions. One such

language is Romanian, as ill ustrated in (63).

(63) a. Cine ce precede?

who what precedes

‘Who precedes what?’

b. *Cine precede ce?

26Franks and Progovac also offer an alternative analysis on which pronominal cliti cs left-adjoin

to je and right-adjoin to all other auxili aries. Needless to say, this analysis is very stipulative.

27The mechanism of pronunciation of lower copies motivated by PF considerations is very

different from PI (for relevant discussion, see Boškovi , 2001a, especially p. 172). It is obviously

very different from it theoretically. The two mechanisms also differ empirically. Thus, PI affects

only PF adjacent elements, which is not the case with the mechanism of pronunciation of lower

copies. However, the application of the latter mechanism depends on the presence of

copies/traces, which is not the case with PI. As a result, none of the constructions that were

argued in section 1.2 to be a problem for PI raise a problem for the pronunciation of lower copies

analysis, discussed in more detail below.

27

- . / 0 1 0 2 3 4 5 . 6 5 0 2 1 0 7 8 9 : . ; < . 1 8 = > ? @ @ @ 6 3 ? @ @ ? 6 A 3 B C 0 5 0 D . 9 7 / C-phrase does not move in

Romanian if it is homophonous with the first fronted wh-phrase.

(64) Ce precede ce?

what precedes what

(65) *Ce ce precede?

Following a proposal concerning Bulgarian made in Billi ngs and Rudin (1996), I propose in : . ; < . 1 8 = (2000b, 2002b) that Romanian has a low-level PF constraint against consecutive

homophonous wh-E C 2 4 5 0 5 3 / C 8 D C 2 F G 0 5 . F B > H I A J > K 9 : . ; < . 1 8 = ? @ @ @ 6 3 ? @ @ ? 6 K 5 C . / B C 4 B B C 0 5 4 L 0holds for a number of Slavic multiple wh-fronting languages.) What about (64)? Given that there

is a syntactic requirement that forces all wh-phrases to move overtly in Romanian (I argue that

the requirement involves focalization), the second wh-phrase should also be moving in the

syntax. Example (64) should then have the S-Structure in (66). (I am ignoring copies of the first

wh-phrase.)

(66) Ce ce precede ce?

If, as we normally do, we pronounce the highest copy of the second wh-phrase in (66), a PF

violation obtains: we end up with a sequence of homophonous wh-phrases. This is precisely the

situation where we are allowed to pronounce a lower copy under Franks’s approach to the

pronunciation of non-B 2 8 1 8 4 G D C 4 8 9 5 J K 4 2 M F 0 8 9 : . ; < . 1 8 = > ? @ @ @ 6 3 ? @ @ ? 6 A B C 4 B B C 8 5 8 5 0 N 4 D B G O / C 4 B

happens in (66).

(67) Ce ce precede ce?

This analysis enables us to derive (64) and account for the contrast between (64) and (63b)

without violating the syntactic requirement that forces all wh-phrases to move overtly in

Romanian, without look-ahead from the syntax to the phonology, and without any PF movement.

There is independent evidence that the second ce in (64) has indeed moved in the syntax.

Thus, it can license a parasitic gap (see (68)), which, as is well -known, can only be licensed by

overt movement. In this respect, the “ce-in-situ” patterns with what in (69), rather than what in

(70), as expected under the proposed analysis.

28

P Q R S T U V W U X U Y U X U Z [ \ [ ] [ ^ _ ` a b c _ d e f e g

what precedes what without subj.particle influence.3p.sg

‘What precedes what without influencing?’

(69) What did John read without fili ng?

(70) *Who read what without fili ng?

Based on this and other arguments given in the works cited above, I will assume that the tail of a

chain can indeed be pronounced instead of the head of a chain iff a PF condition requires it.28

Returning now to je, our job is twofold. First, we need to account for the fact that, in spite

of following them in the phonology, je is higher than pronominal cliti cs in the syntax. Second,

we need to account for the fact that je is higher in the structure in the syntax when it occurs alone

than when it occurs with a pronominal cliti c, as indicated by the contrast between (58a) and

(58b). In (58a), where je is the only cliti c, je is higher than the adverb even on the sentential

reading of the adverb. This is not the case in (58b), where je cooccurs with a pronominal cliti c. A

related fact to be accounted for is that je follows clause-mate pronominal cliti cs only when it is

in a cliti c cluster, as the contrast between (54) and (57) with respect to the linear order of je and

ga shows.

Before accounting for the dual behavior of je we need to address the issue of the structure

of periphrastic constructions in SC. Recall that, as discussed in section 1.3., pronominal cliti cs 28Notice that in the pronounce a copy analysis, (ia) and (37a), repeated here as (ib), can both be

analyzed as having the reflexive cliti c above the participle. A lower copy of the reflexive cliti c is

pronounced in (ib) in order to avoid having a reflexive cliti c immediately follow an I-phrase

boundary, which follows rekla.

(i) a. Oni su se predstavili Petru.

They are self.acc introduced Petar.dat

‘They introduced themselves to Petar.’

b. Oni su, kao što sam vam rekla, se predstavili se Petru.

they are as am you.dat said introduced self.acc Petar.dat

‘They, as I told you, introduced themselves to Petar.’

As noted in fn. 15, the parenthetical can also follow the highest copy of the reflexive, as in Oni

su se, kao što sam vam rekla, predstavili Petru.

29

are hierarchically arranged in different maximal projections in the syntax. More precisely, they

are located overtly in their Case-checking Agro projections. Den Dikken (1994) argues that in

SC constructions involving the auxili ary verb be, object agreement projections are generated

above the VP headed by be. Given that in the syntax all clitic forms of the auxili ary be precede

pronominal cliti cs, it must be the case that the auxiliary undergoes overt movement to a head

position above pronominal cliti cs, which was suggested above to be the highest head in the spli t

I, namely, Agrs. Constructions involving a dative and an accusative cliti c as well as the auxili ary

cliti c je then abstractly have the S-Structure in (71).

(71) jei [AgrioP dative cliti c [AgrdoP accusative cliti c [VP/AuxP jei ...]]]

Notice that a copy of je is present both above and below pronominal cliti cs. Suppose now that

there is a low level constraint on the final PF representation requiring that in a cliti c cluster (i.e. a

sequence of two or more cliti cs) je must follow all other cliti cs.29 The constraint would force the

pronunciation of je in the tail of the chain created by its movement in (71). Since the

pronunciation of je in the head of the chain would lead to a PF violation, pronunciation in the tail

of the chain is sanctioned, in fact, required.30 Since on this analysis, je is higher than pronominal

cliti cs in the syntax, the data in (55)-(58) can be easily accounted for. (The PF constraint in

question is irrelevant in (55)-(57) since je is not a part of a cliti c cluster in the final PF

representation. Therefore, the highest copy of je is pronounced.) Notice also that the fact that je

can precede a subject-oriented adverb only when it does not follow a pronominal cliti c (see (58a-

b)) is also accounted for. The reason why je exhibits different behavior with respect to structural

height when it occurs alone and when it cooccurs with a pronominal clit ic is that in the former

case we are pronouncing the highest copy of je, which is higher than sentential adverbs, while in

the latter case we are pronouncing a lower copy of je, which is lower than sentential adverbs.

The dual behavior of je with respect to pronominal cliti cs--je precedes pronominal cliti cs, i.e., it

29I am using here the traditional term cliti c cluster for ease of exposition. What I really have in

mind is the cliti c group, which is a prosodic unit. I am not positing here any kind of syntactic

clustering of cliti cs. It is important to bear this in mind. Below I give a more precise formulation

of the requirement in question. I also provide principled motivation for the requirement.

30h i j k l m k n o p q r s s t u v t r w x h y z { { | y } } ~ u } } ~ | | � � | � } o k i u � k � � | � o i { k � } ~ | � | � � | � o n | � � o } o � se, which

follows other pronominal cliti cs, can be accounted for in a similar way. As noted in this work,

this analysis explains the well -known je-drop in the presence of se. (Other auxili ary cliti cs

cannot be dropped in the presence of se� x � | | � ~ k � | n | � � j k l m k n o p q r s s t u v � s x u i � � | � | � | i � | ytherein for another approach to se.

30

is higher than pronominal clit ics, in the syntax, but follows them in the final PF representation--

as well as the dual behavior of je with respect to structural height--it is higher in the structure

when it occurs alone than when it cooccurs with a pronominal cliti c, is thus accounted for.31

Furthermore, this is done without positing any PF movement.

A question arises now what the source of the PF requirement on je � � � � � � � � � � � � � � � � � � �� � � � � � � � �   ¡ � � ¢ £ ¤ ¥ ¦ § � � � � � ¨ � � � � © ª � « � � � � � � ¬ � � ­ � � � � � ¦ ¤ � ¦ je is in the process of losing its

cliti chood with respect to a number of phenomena, where je behaves differently from other

auxili ary cliti cs. (However, je has not completely ceased to be a cliti c, so it still cannot occur

sentence initially. In other words, it is still subject to (5).)32 It seems plausible that this should

lead to placing je at the very edge of the cliti c cluster, given that non-cliti c material that does not

form a separate I-phrase (and je clearly cannot form an I-phrase on its own) cannot intervene

between cliti cs in a cliti c cluster in SC, as discussed in section 1.1. We would then expect a

development of a low level constraint that would force je to be located either in the initial or the

final position of the clit ic cluster. We can assume that the final position is chosen arbitrarily.

However, we may be able to do better than that. If, following a suggestion by Klaus Abels

(personal communication), we assume that as a result of being in the process of losing its

cliti chood, je does not allow cliti cization across it but is not strong enough to serve as a cliti c

host itself, we would be forced to pronounce je following all other cliti cs. The peculiar

requirement that je follows all other cliti cs is thus explained.

There is another way to force je to be located in the cluster final, rather than the cluster

initial position. Under the current analysis, the only way to place je at the edge of a cliti c cluster

is to pronounce one of the members of the chain created by the movement of je at the edge of the

cluster. This can be easily accomplished for all cases by pronouncing the tail of the chain, since

the tail i s always located lower than other cliti cs. The desired result, however, cannot be

consistently achieved by pronouncing the head of the chain created by the movement of je since

the head of the chain is located lower than the question cliti c li , as the following construction,

where the auxili ary cliti c follows li , shows.33

31An anonymous reviewer observes that under the morphological template analyses je is also

required to follow all cliti cs in a cliti c cluster. However, in contrast to the current analysis, these

analyses leave the dual behavior of je with respect to the syntax and phonology unexplained.

32Alternatively, it is possible that je has not yet fully gained its cliti c properties. The analysis ® � � � � ­ � ¯ � � � ¦ � � � � ° � � « � � � � � � � � � � � � � � � � £ � � ­ � � � � � ¯ ± � ² ¦ � � © � © ¦ � ¦ ¤ � � � � � � � 33This way (i.e. by pronouncing the head of the je-movement chain) the desired result can be

achieved for some but not all � � ¯ � � � � ¦ £ � � � � � ª ¤ � � � � © � � � � � ¯ � � � � ° � � � � © ¦ � � � � � � � � � � � � � � � ³ � ´ � �

31

(72) Kola li si mu dao?

car Q are him.dat given

‘Was it a car that you gave him?’

I conclude therefore that the PF requirement on je proposed above can be explained in a

principled way. The requirement has developed as a consequence of je losing its cliti chood. It’s

task is to pull je to the edge of the cliti c cluster so that (5) can be satisfied in cliti c clusters

involving je.

I now turn to additional data involving jeµ ¶ · ¸ ¹ º » ¼ · ½ º º · ¾ » ¿ À Á Â ¾ ¸ Ã Ä Â Å Â Ã Æ ¶ » Ã Ç · ¹ Â Ècommu¹ ¸ É Â º ¸ · ¹ Ê µ Ë Ì ¸ É Ì Ë » Ã » ¹ · º ¼ ¸ Ç É ½ Ç Ç » ¼ ¸ ¹ Í · Î Ï · Å ¸ Ð Æ Ñ Ò Ò Ó Â Ê Ô Õ Ë ¸ È È Ç Ì · Ë º Ì Â º º Ì » ¼ Â º Â ¸ ¹question can also be straightforwardly accounted for under the above analysis of je. They will

furthermore give us an additional insight into the derivation of periphrastic constructions.

Ö · ¹ Ç ¸ ¼ » Ã × ¸ Ã Ç º º Ì » × · È È · Ë ¸ ¹ Ø ¼ Â º Â ¹ · º » ¼ ¿ À Ä Â Å Â Ã Ô

(73) a. [Dao ga Mariji ] je Ivan.

given it.acc Marija.dat is Ivan

‘Give it to Marija, Ivan did.’

b. [Dali ga Mariji ] su Ivan i Stipe.

given it.acc Marija.dat are Ivan and Stipe

‘Give it to Marija, Ivan and Stipe did.’

Both constructions in (73) involve fronting of a phrase below the auxili ary cliti c. We can see that

je behaves like other auxiliary cliti cs in that the fronted phrase, located somewhere in the

complement of the auxiliary cliti c, can contain a pronominal cliti c. There are several ways of

analyzing (73). One possibili ty is to analyze it as involving AgroP fronting. The constructions

and references therein for arguments that the initial je in (i), which contrasts with (ii ), is not the

auxili ary cliti c je. Rather, jeli is a non-cliti c counterpart of the cliti c li (see also Browne 1975, Ù Â ¼ Â ¹ · Å ¸ Ð -Ú · É ¸ Ð Ó Û Ü Ü Ý Þ ß -Þ Û µ Â ¹ ¼ à · ¾ ¸ Ð Ó Û Û á Ê Ô

(i) Je li (je) on istukao Petra?

Q is he beaten Petar

‘Did he beat Petar?’

(ii ) *Si li ti istukao Petra?

are Q you beaten Petar

‘Did you beat Petar?’

32

will t hen involve lower pronunciation of the pronominal cliti c within VP, which is necessary to

satisfy the encliti c requirement on SC clitics, plausibly a PF requirement. (Recall that a lower

copy can be pronounced if this is necessary to avoid a PF violation.)34

(74) a. [AgroP ga dao ga Mariji ] je Ivan.

b. [AgroP ga dali ga Mariji ] su Ivan i Stipe.

â ã ä å ã æ ç è æ é å ê ë ì í î ç ïavar (personal communication), who provided the following data, a very

interesting contrast obtains when in a double object construction we attempt to pronounce a

pronominal cliti c with the auxili ary. This cannot be done in a construction with je, but can be

done with other auxili aries.35

34ð ñ æ ç æ ì ç æ ò ó ä ì ô ò æ ç õ ì ò î è æ ã ò ä ò ñ î ã ì õ ì ô ê ã î ã æ ö ÷ ô ä ç æ é î õ ø ä ù ú ä è î û ü ý þ þ ÿ ì � � ñ ì ÷ ò æ ç � � ò ñ ì ò ó ä � ô éallow pronunciation of the pronominal cliti c in AgroP. One possibili ty is that both the participle

and the pronominal cliti c head move to Agro. (Assuming that the clit ic is a non-branching

element, in Chomsky’s 1995 system it would be ambiguous between a phrase and a head and

could therefore move to either SpecAgroP or Agro.) Both the participle and the cliti c could then

be located in their highest position in Agro in (73)ü ã æ æ ø ä ù ú ä è î û ý þ þ ÿ ì � ä ç é æ ò ì î ô ã ä � ò ñ æ

ì õ ì ô ê ã î ã � � â õ ä ò ñ æ ç ÷ ä ã ã î å î ô î ò ê æ ö ÷ ô ä ç æ é î õ ø ä ù ú ä è î û ü ý þ þ ÿ ì � î ã ò ñ ì ò ò ñ æ ÷ ì ç ò î � î ÷ ô æ í ä è æ ã ò ä ìparticipial aff ix head Part0

� ó ñ î � ñ ò ì ú æ ã â � ç ä � ì ã î ò ã � ä í ÷ ô æ í æ õ ò ü ã æ æ ì ô ã ä ø ä ù ú ä è î û 1997a for

relevant discussion). On this analysis, (73) would involve PartP fronting and the pronounced

copy of the cliti c could be located in SpecAgroP. Notice, however, that the participle movement

to Part/Agro could not be obligatory given the discussion of (76) and (78) below.

35ï ì è ì ç é ä æ ã õ ä ò � î è æ[Dao joj knjigu] je Ivan, which is expected to be acceptable given the

grammaticali ty of (73a). Notice also that the contrast between je and other auxili aries also

obtains if the accusative is left behind.

(i) a. *[Dao Mariji ] ga je Ivan.

given Marija.dat it.acc is Ivan

b. ?[Dali Marij i] su ga Ivan i Stipe.

given Marija.dat are it.acc Ivan and Stipe

‘Give it to Marija, Ivan and Stipe did.’

The relative grammaticali ty of (ib) is not surprising either if lexical dative NPs do not have to

undergo overt object shift or if double object constructions involving a lexical NP correspond to,

or at least can correspond to, the English to-phrase double object construction, where the theme

33

(75) [Dao Mariji knjigu] je Ivan.

given Marija.dat book.acc is Ivan

‘Give Marija the book, Ivan did.’

(76) * [Dao knjigu] joj je Ivan.

given book.acc her.dat is Ivan

(77) [Dali joj knjigu] su Ivan i Stipe.

given her.dat book.acc are Ivan and Stipe

(78) ?[Dali knjigu] su joj Ivan i Stipe.

given book.acc are her.dat Ivan and Stipe

Under the arbitrary morphological template approach to cliti c ordering, the ungrammaticali ty of

(76) is surprising in light of the acceptabili ty of (78). There is nothing in this approach, where the

order of cliti cs within the cliti c cluster is simply stipulated, that could give us the contrast � � � � � � � � � � � � � � � � � � � � � � � � � � � ! � " � � � # � $ � � � % & & & ' ( % �-330), and Franks and

Progovac’s (1994) analyses, which in order to account for (54) assume that je is lower in the

structure than pronominal cliti cs, also fail to account for the contrast in question. (Recall that

these analyses also fail to account for the data in (55)-(58), which show that je is higher than

pronominal cliti cs in the syntax.) The same holds for the Franks and Progovac suggestion from

fn. 26. So, how can we account for the ungrammaticali ty of (76) given that, as indicated by (78),

the relevant fronting operation is possible in the syntax?36

The current account of the je-final effect makes it possible to account for the mysterious

contrast between (76) and (78) in a principled way. Recall that under the current analysis,

auxili ary je constructions and constructions involving other cliti c forms of biti ‘be’ do not differ

at all i n their syntactic derivation. The only difference between the two concerns PF. In the

presence of a PF-adjacent pronominal cliti c, we always pronounce a lower copy of je. With other

auxili ary cliti cs, if there are no other interfering factors, the highest copy of the auxili ary is

pronounced even in the presence of a pronominal cliti c. The contrast between (76) and (78) can

then be straightforwardly accounted if there is no copy of the auxili ary cliti c following the

is higher in the structure than the goal. Having free goal theme/theme goal order in double object ) * + , - . / ) - 0 * + , 0 + 1 * 2 1 0 + 3 4 2 5 6 0 ) 4 2 7 8 9 4 , : . * : * , 5 ; 0 + < - = 5 : 4 + * 1 0 > ? @ A A A B ? , 5 5 4 2 , * C 0 D 4 3 4 E 41997 for Japanese), may also be a possibili ty. It is worth noting here that the application of

height tests to double object constructions with lexical NPs does not give completely clear results 0 + < F ? , 5 5 < - = 5 : 4 + * 1 0 > @ A A A B G 36Needless to say, *Dao knjigu je joj Ivan is also unacceptable.

34

pronominal cliti c in the remnant of the VP fronting in the constructions in questions.37 This

assumption makes (76) underivable without affecting the acceptabili ty of (78). On the other

hand, accomplishing this seems to be impossible under the morphological template approach,

where the two constructions should have the same status.

Let us now fill i n the details of the analysis. The first question to address concerns the

status of the phrase undergoing fronting in (76) and (78). One possibilit y is that the phrase is

what is labeled as VP/AuxP in (71), i.e. the phrase where the auxili ary is base-generated. Both

the auxili ary and the clit ic would undergo movement out of that phrase prior to the fronting. The

auxili ary, which morphologically agrees only with the subject, would move to Tense and then to

Agrs; it would not move through Agro phrases.38 Where is the subject located? We may be

dealing here with the notorious issue of the position of postverbal subjects in pro-drop languages,

resolving which goes well beyond the scope of this paper. I will simply locate the subject in a

phrase whose label I leave open.39 I indicate only the traces of the cliti cs. (The participle could

actually be moving to the auxili ary, which would be followed by auxil iary excorporation; see H I J K I L M N O P P Q R S T U V Q P W T X M L Y Z [ Y \ Zructure for (78). Example (76) has the same structure.)

(79) [VP/AuxP ti [VP Dali t j knjigu]] [AgrsP sui [TP ti [AgroP joj j [XP Ivan i Stipe]] ]]

There is only one issue left to be addressed. Given the structure in (79) it is clear why (76) is

unacceptable. There is no copy of the auxili ary following the pronominal cliti c, hence the

auxili ary cannot be pronounced following the pronominal cliti c. Could we pronounce the

auxili ary in the position of its original trace in Aux? Given the structure in (79), this should lead 37] Y ^ R _ _ Z [ R Z ` I _ _ I a M U X H I J K I L M N V b c c O R W T R \ \ d e Y Z [ R Z Z [ Y f Y M \ U I g h e I L Y e Y U Z S i [ Y I U _ jrelevant way for PF considerations to affect word order is by licensing pronunciation of a lower

copy.

38Technically, the analysis violates the Head Movement Constraint. However, following standard

assumptions, I assume that the constraint has no status in the grammar.

39There are several possibiliti es concerning how the subject would get to this position. Given that

SC is a heavily scrambling language, we could be dealing here with scrambling. SpecAgrsP

could be either empty or fill ed by a null expletive. Or the subject could actually be moving to

SpecAgrsP through XP, a lower copy of the subject being pronounced to avoid violating the \ Y ^ I U k l I \ M Z M I U f Y m d M f Y e Y U Z I U ^ _ M Z M ^ \ V \ Y Y H I J K I L M N b c c O R ` I f \ Y L Y f R _ ^ R \ Y \ a [ Y f Y Z [ Y \ d n o Y ^ Z M \pronounced in a lower position for this reason). Another possibili ty is right dislocation or, more

generally, base generation of the subject in the position in question.

35

to a violation of the encliti c requirement on the auxili ary cliti c (cf. *Je dao knjigu joj Ivan). p q r s t s u v w x v y z z { | | s z } s ~ w � � q � � q t w � � � � � � y � v } � s � y u } w � w � � s y � } { y � � � y ~ � q w � z } q } � s y { � w � w y u � w �Aux, after which the auxili ary excorporates to move to a higher head position, we could end up

with *Dao je knjigu joj Ivan, where the encliti c requirement on the auxili ary cliti c is satisfied.

Can this derivation be blocked? There are several rather straightforward ways to block it. One

possibili ty is to assume that, as in English, I (i.e. Agrs/T) must be lexicalized in constructions

involving VP fronting.40 Recall also that the constraint that normally li censes lower

pronunciation of je in the presence of a pronominal cliti c requires that je be pronounced at the

right edge of the cliti c cluster. Let us assume that this is the proper formulation of the constraint:

Pronounce je at the right edge of a cliti c cluster. (Recall that cliti c cluster is a sequence of more

than one cliti c.) The constraint can only li cense lower pronunciation of je if this wil l lead to the

pronunciation of je at the right edge of the cliti c cluster. This, however, is not the case with the

pronunciation of je in Aux. Hence, the constraint in question cannot license this pronunciation.41

40Taking the requirement literally would actually block the pronunciation of the auxili ary

following the pronominal cliti c even if there were a copy of the auxili ary cliti c following the

pronominal cliti c.

41Another possibili ty is to appeal to a suggez } w q � � y ~ s w � � q � � q t w � � � � � � y � x q u q } � s u � q { } �

Slavic languages that cliti cs belonging to the same I-phrase must be parsed into a prosodic

constituent which attaches to its host as a unit. The requirement obviously cannot be satisfied if

the auxili ary clit ic is pronounced in the fronted VP in (79). This analysis makes an interesting

prediction. Notice that in all the cases above where clause-mate cliti cs are not adjacent and

therefore cannot form a prosodic constituent, the cliti cs are not located in the same I-phrase. As ~ w z � { z z s ~ w � � q � � q t w � � � � � � y � v y x u q � } s ~ � � w � � � � y � � s v � { } ~ q s z � q } � y t s } q � s v � y u z s ~ y z yseparate I-phrase. The option of parsing the fronted VP as a separate I-phrase, as a result of

which the auxiliary cliti c pronounced in the fronted VP and the pronominal cliti c in AgroP

would not be I-phrase-mates, cannot be taken in (79) since this would lead to an I-phrase

boundary intervening between the pronominal cliti c and its host, which is disallowed (see � q � � q t w � � � � � y � � p q r s t s u v } � s � u q � � s � ~ q s z � q } y u w z s w � � w � v r � s u s } � s x u q � } s ~ � � w z � y u z s ~ y z yseparate I-phrase. (An adverb is added because constructions with sentence final cliti cs are

sometimes disfavored.)

(i) [Dao je knjigu]# Ivan joj (� { � s

).

given is book.acc Ivan her.dat yesterday

Under the analysis suggested in this footnote we may then expect (i) to be acceptable. This is in

36

In conclusion, the pronounce-a-lower-copy analysis of the exceptional behavior of je

provides us with an account of the dual behavior of je with respect to pronominal cliti cs in the

syntax and phonology, more precisely, the fact that je precedes (i.e. is higher than) pronominal

cliti cs in the syntax, but follows them in the phonology. The analysis also accounts for the dual

behavior of je with respect to structural height, more precisely, the fact the je is higher in the

structure when it occurs alone than when it cooccurs with a pronominal cliti c. Both of these are

accomplished without positing any PF movement. PF is allowed to affect word order, but only

through licensing pronunciation of lower copies, not through actual PF movement. The analysis

also gives us an account of the otherwise mysterious contrast between je and other auxiliaries

with respect to leaving a pronominal cliti c behind with the auxiliary in constructions involving

VP fronting. The analysis of this contrast has provided us with several insights into the

derivation of auxili ary+participle constructions.

2. Bulgarian and Macedonian clitics

Bulgarian and Macedonian cliti cs occur adjacent to the verb in finite clauses like (80) (cf.(80d)).

In other words, they are verbal rather than second position cliti cs in the context in question.42

contrast to (ii ), where the lower pronunciation of the auxili ary cliti c in the fronted VP is not

required, hence disallowed.

(ii ) [Dali su knjigu]# Ivan i Stipe joj � � � � � � . given are book.acc Ivan and Stipe her.dat yesterday

Unfortunately, not having access to speakers who allow VP fronting in constructions with

auxili ary cliti cs, I am unable to check the prediction of the analysis suggested in this footnote.

(The alternative analyses given in the text would predict both (i) and (ii ) to be bad.)

42The situation is more complicated in certain non-finite clauses in Macedonian, where the

second position cliti c pattern emerges. I ignore these contexts here. For relevant discussion, see   ¡ ¢ £ ¡ ¤ ¥ ¦ � § ¨ ¨ © ª � « ¬ ª ¥ ­ £ � © ® ® ¯ � « ° ± ª ­ £ ² � © ® ® ¯ « § ¨ ¨ ¨ ª � « ° ± ª ­ £ ² ª ­ ³ ´ ¥ ­ µ � § ¨ ¨ ¨ � « ¶ � µ � ­ ³ ± �(1998, 1999), · ¸ ¹ º » ¼ ½ ¾ ¿ À Á Á Â Ã À Á Á Ä Å Ã · ¼ » ¸ Æ » Ç È É Ê Ë Ì Í » Ç ½ Î É Ç È · Ç ½ ¸ · Ï Ð ½ Ñ ½ · Ê Ò Ó Ô · Ê Ç ½ Î ½ Ô Ñ Éconstructions, in constructions with cliti c auxili aries the participle plays the same role as the

finite verb in (80) with respect to adjacency, and the auxili ary does when the auxil iary itself is a

non-cliti c. Some (though not all ) Bulgarian speakers do allow some short adverbs to intervene

between cliti cs and the finite verb (see, for example, Krapova 1997, 1999). However, as

discussed in Boškovi ¾ (2001a:181), the adverbs in question incorporate into the verb and thus

37

(80) a. Petko mi go Õ Ö Õ × Ø Ù × Ú Ö Û B: OK Mac: OK

Petko me.dat it.acc gave yesterday

‘Petko gave me it yesterday.’ Ü Û Ý Ù × Ú Ö Þ × ß à á mi go dade B: OK Mac: OK â Û Ý Ù × Ú Ö mi go dade Petko. B: OK Mac: OK

d. Petko mi go Ø Ù × Ú Ö Õ Ö Õ × Û B: * Mac: *

e. Mi go Õ Ö Õ × Þ × ß à á Ø Ù × Ú Ö Û B: * Mac: OK

f. Dade mi go Þ × ß à á Ø Ù × Ú Ö Û B: OK Mac: *

g. ã Ö Õ × Ø Ù × Ú Ö ä å æ á Þ × ß à á Û B: * Mac: * ç Û Ù × è Õ × à Ö Þ × ß à á mi go dÖ Õ × Ø Ù × Ú Ö Û B: OK Mac: OK

that Petko me.dat it.acc gave yesterday

The grammaticali ty of (80h), which is unacceptable in SC, and the fact that the adverb in (80b)

does not have to be followed by a pause, in contrast to its SC counterpart, indicate that Bulgarian

and Macedonian cliti cs are not second position cliti cs. The contrast between Bulgarian and

Macedonian (80e) indicates that Bulgarian cliti cs must encli ticize, whereas Macedonian cliti cs

can procliti cize. Macedonian cliti cs always precede the verb in the context in question. Bulgarian

cliti cs precede the verb unless preceding it would result in a violation of their encliti c

Ú × é ê å Ú × ä × ë ß Û ì ë ß ç Ö ß â Ö í × ß ç × î ï á ð ð á ñ ß ç × Ø × Ú Ü Û ò í Õ å í â ê í í × Õ å ë ó á ô à á Ø å õ ö ÷ ø ø ù Ö ú û ß ç å í í ß Ö ß × á ïaffairs can be straightforwardly accounted for under Franks’s (1998) copy and delete approach to

the pronunciation of non-trivial chains, given that a copy of pronominal cliti cs is present both

Ö Ü á Ø × Ö ë Õ Ü × ð á ñ ß ç × Ø × Ú Ü ö í × × ó á ô à á Ø å õ ÷ ø ø ù Ö ï á Ú Õ å í â ê í í å á ë á ï ß ç × ü Ú × â å í × ü osition of these

copies). Recall that under the copy and delete approach, the tail of a non-trivial chain is

pronounced instead of the head iff the pronunciation of the tail of the chain is necessary to satisfy

a PF requirement. This approach straightforwardly captures the generalization that the verb can

precede a cliti c in Bulgarian only when no other lexical material is located in front of the cli tic.

(Notice the ungrammaticali ty of (80g).) Only in this situation will we be able to pronounce the

lower copy of the cliti c, which is located below the verb. If there is lexical material preceding the

end up being part of the cliti cs+V cluster. A word of caution is in order regarding using e ‘ is’ as

the only cliti c in testing V-adjacency since, as noted above with respect to SC je, e is not fully a

cliti c. (It does not always pattern with other cliti cs.) For more general discussion of clausal

cliti cization in Bulgarian and Macedonian, see Alexander (1994), Alexandrova (1997), ý þ ÿ � � � � � � þ � � � � � � � � � � � � � � � � � � � � þ � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � þ �-

Vulchanova (1995), Dimitrova-Vulchanova and Hellan (1999), Ewen (1979), Franks (1998), � � � � ! � � " # $ � % & ' ( ( ( ) * + , % , � " � , & - . . . * ' ( ( ( ) * / , � 0 , 1 & - . . 2 ) * 3 4 5 � 6 % , & - . 7 8 ) * 3 $ 1 , � 9& - . . 7 ) * 3 6 " $ � & - . . 7 ) * : ; < $ ; & ' ( ( ( ) * = 9 > $ ? & - . . 8 * - . . 7 * ' ( ( ( ) * � > 9 � % 9 @ < , � ! A

38

cliti c in its raised position, the head of the chain of cliti c movement has to be pronounced.

(81) a. X cliti c V cliti c

b. cliti c V cliti c

Since in Macedonian nothing goes wrong in PF if we pronounce the head of the cliti c chain, we

always have to pronounce the head of the cliti c chain, located above the verb. As a result, the V-

cliti c order is underivable in Macedonian.

(82) (X) cliti c V cliti c

The contrast in the acceptabili ty of (80e-f) in Bulgarian and Macedonian, as well as the role of

phonology in the possibili ty of the V-cl order in Bulgarian, are thus straightforwardly captured

without appealing to any kind of PF movement.

It is standardly assumed that Bulgarian and Macedonian cliti cs cluster together with the

verb in the same head position. There are two reasons for this. First, it is much harder to split the

cliti cs+V cluster, which I will refer to as the extended cliti c cluster (ECC), in these languages

than in SC, as ill ustrated by the ungrammaticali ty of (80d), which is acceptable in SC, as shown

in (83).

(83) Jovan mi ga B C D E F G F E H (SC)

Jovan me.dat it.acc yesterday gave

‘Jovan gave it to me yesterday.’

As an additional ill ustration of this fact I give the Bulgarian elli psis examples in (84), which are

G I I E J K G L M E N O P Q H R P E E R S T G U G O F R V W G U U H P E E G M X Y Z Y [ \ Y ] N ^ S _ _ W G ` Y a b Y a E a E M E ] G O K F N X I C X X N Y O Y `Bulgarian and Macedonian.)

(84) a. *Nie sme mu go dali i vie ste mu go dali (sc što). (B)

we are him.dat it.acc given and you are him.dat it.acc given (too)

b. *Nie sme mu go dali i vie ste mu go dali (sc što).

Second, the verb carries cliti cs along when moving to a higher head position in the languages in

question. This is ill ustrated by the following li -constructions from Macedonian, where the verb

moves to li , standardly assumed to be a Q-marker, carrying the whole cliti c cluster (ne si mu gi in

39

(85a) and ne mi go in (85b)) along.43

(85) a. [Ne si mu gi dal] i li ti parite? (Mac)

neg are him.dat them.acc given Q the-money

‘Haven’ t you given him the money?’

b. [Ne mi go dade] i li ti vd era?

neg me.dat it.acc gave Q yesterday

‘Didn’ t he/she/you give it to me yesterday?’

In light of this, consider the following construction.

(86) Ti ne si mu gi dal. (B/Mac)

you neg are him.dat them.acc given

‘You haven’ t given them to him.’

The following is the standard analysis of (86) (see, for example, Franks 1998, Franks and King e f f f g h i j k l m n n o g h i j k l g p q r s t q g u k v v k l w x g r l j u r t q s r l m n n n g r l j y z s k { m n n | } ~ne si mu gi

dal are all l ocated in separate projections. The verb moves up through successive cyclic

rightward adjunctions so that in the end we end up with the ne si mu gi dal cluster, with the order

within the cluster mirroring the structural height of the relevant elements prior to cluster

formation. The reason for assuming that the ECC formation proceeds through rightward

adjunction is that the structural height of relevant elements prior to the ECC formation

corresponds to the left-to-right order of heads within the ECC. Rightward adjunction preserves

the order.

The above derivation is obviously incompatible with Kayne (1994), which disallows

rightward movement. In fact, it is a tacit assumption in the literature on South Slavic cliti cs that

Kayne’s Linear Correspondence Axiom (LCA) cannot be maintained, at least not for head

43Bulgarian li -constructions involve an interfering factor in that PF requirements on cliti cs in

some cases force pronunciation of lower copies of some cliti cs. However, it is shown in u z � � z � k { � e f01a,b) (see also Rudin, Kramer, Billi ngs, and Baerman 1999) that in Bulgarian, the

ECC also moves as a unit to li in the syntax. For discussion of Bulgarian and Macedonian li , see

also Rivero (1993), Franks (1998), Franks and King (2000), Izvorski, King, Rudin (1997), King � m n n | } g h i j k l � m n n � } g h i j k l g p k l w r l j � � � z q x � k � m n n � } g r l j y z s k { � m n n | } g r s z l w z � � t q x � � t tr v x z u z � � z � k { � e f f m r } � z q r l r l r v � x k x z �

li that is fully consistent with the cliti cs-as-non-

branching elements hypothesis, discussed below.

40

movement (see in this respect Chomsky 1995, who adopts the gist of Kayne’s system but leaves

open the possibili ty that it might not be applicable to head movement, essentially through a

stipulation).

A question that we need to answer, then, is whether Bulgarian and Macedonian ECC can

be formed through leftward instead of rightward head adjunctions while still having the left-to-

right order of elements within the ECC reflect the higher-to-lower hierarchical structure of

relevant elements prior to the ECC formation. At first sight, the answer seems to be no. � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �   ¡ ¢ � � £ ¢ � £ £ ¢ � � � � ¤ £ ¥ � ¦ ¦ § � ¡ � ¨ � � � ¤ � ¨ ¦ � © � � § � ªaccomplishing this which resolves a potentially very serious problem for Kayne’s (1994) system.

More precisely, I show that given economy of derivation, the task at hand can be accomplished if

we take seriously Chomsky’s (1994) suggestion that cliti cs are non-branching elements.

Chomsky (1994) proposes a phrase-structure system that allows for the existence of

elements that are at the same time phrases and heads, the prerequisite for the ambiguous XP/X0

status of an element X being that X does not branch. (In fact, every non-branching element is

automatically both a phrase and a head in Chomsky’s 1994 system.) Chomsky mentions cliti cs as

a possible example of such ambiguous XP/X0 � ¦ � « � � £ ¡ ¬ � � � � � � � � � � ­ ­ ® � � ¨ � � � � © � ¡ � « ¨ � � � ¤ � ¦evidence for this suggestion, which can be interpreted as a way of capturing the intuition that

cliti cs have less structure than their non-cliti c counterparts (assuming that the latter do branch), a

position argued for convincingly in Cardinaletti and Starke (1999). Suppose now that cliti cs are

indeed ambiguous XP/X0 elements, which means that they do not branch. (This would be

necessary but not suff icient for something to be a cliti c.) This proposal has an interesting

consequence for auxili ary cliti cs. Auxili ary cliti cs such as the one in Bulgarian (87a) can no

longer be analyzed as the head of a phrase taking another phrase as its complement, as shown in

(87b). Instead, we can analyze the XP as headed by a null element, with the auxili ary cliti c being

located in its specifier, as shown in (87c). Since X rather than the auxiliary cliti c is taking a

complement, the cliti c remains non-branching and, therefore, an ambiguous XP/X0 element.44

(87) a. Petko e � « � � � ¦ � ° � � � ¬

Petko is left yesterday

‘Petko left yesterday.’

b. Petko [XP [X’ e ± ¯ � « � � � ¦ � ° � � � ² ² ²

c. Petko [XP e [X’ ³ ± ¯ � « � � � ¦ � ° � � � ² ² ²

44The analysis can also be straightforwardly extended to auxili ary cliti cs in SC, discussed in

section 1.

41

The negative cliti c also can no longer be analyzed as the head of a phrase taking another phrase

as its complement. In accordance with the cliti cs-as-non-branching-elements hypothesis, it is

generated as a specifier of a null head, which takes a phrasal complement. Example (86) then has

the following S-Structure prior to the ECC formation.45

(88) Ti [NegP ne [Neg’ [AuxP si [Aux’ [VP mu [V’ dal gi]]] ]]]

Recall that under the standard analysis, the cliti c-verb complex in Bulgarian is formed by right-

adjoining the verb to the cliti cs. In Boš µ ¶ · ¸ ¹ º » » ¼ ½ ¾ º » » º ½ ¿ À Á Â Ã Ã Ä Á Å Å Æ ½ Å Å Æ Ä Ç È · Å · Ç -verb

complex is instead formed by left-adjoining the cliti cs to the verb, which is in accordance with

Kayne’s (1994) system.46 I will demonstrate that the leftward adjunction analysis yields the

45I am assuming Marantz’s (1993) version of Larson’s (1988) VP shell analysis of double object

constructions, where the dative is generated in the Spec and the accusative in the complement

position of the lower VP shell . I ignore Agr phrases, the higher VP shell , and the trace in SpecVP

since they are irrelevant for our current concerns. (As will become obvious during the discussion

below, object cliti cs move overtly to Agro in Bulgarian and Macedonian as part of the ECC.) É Ä Ä ¾ Æ µ Ê Ä ¶ Ä Ë ¾ Ì µ Í ´ µ ¶ · ¸ ¹ º » » ¼ ½ ¿ ¾ Ê Æ Ä Ë Ä · Å · Á Á Æ µ Ê Î Å Æ ½ Å the analysis of (86) about to be given

can be readily restated in a system where the pronominal cliti cs are generated within AgroPs.

(The reader should bear this in mind.) Bulgarian pronominal cliti cs are typically suggested to

originate outside of VP because the language allows cliti c doubling (see, for example, Franks

¼ Ï Ï Ð ¾ Ñ Ë ½ Î ´ Á ½ Î Ò Ó · Î Ã º » » » ¾ Ô Â Ò · Î ¼ Ï Ï Õ ¾ ½ Î Ò Ö µ × · ¸ ¼ Ï Ï Ø ¿ Ù Ú µ Ê Ä ¶ Ä Ë ¾ Å Æ Ä Ë Ä ½ Ë Ä ½ Î Â × Û Ä Ë µ Üsuccessful analyses of cliti c doubling that are consistent with pronominal cliti cs originating

within Ý Þ ¹ Á Ä Ä Ì µ Í ´ µ ¶ · ¸ º » » ¼ ½ ß ¼ Ð Õ ½ Î Ò Ë Ä Ü Ä Ë Ä Î Ç Ä Á Å Æ Ä Ë Ä · Î ¿ Ù 46Following Baker (1988), I assume that a nominal/pronominal element can be Case-licensed by

undergoing head-movement to a Case-licensing head. Notice that being ambiguous XP/X0

elements, cliti cs can undergo head-adjunction. As a technical implementation of the adjunction,

we can assume that the main verb is lexically specified with an Attract All property in the sense

of Boškovi ¸ (1999) for pronominal and auxil iary cliti cs. The verb would then attract all

pronominal and auxili ary cliti cs. In Boškovi ¸ (1999) I show that multiple movement to the same

element as a result of an application of the Attract All mechanism generally results in free

ordering of elements undergoing the movement. However, this would not happen in the case

under consideration as a result of the earliness effect of economy of derivation discussed directly

below.

It is also worth noting here that leftward adjunction of a cliti c to its host seems to be

independently necessary for cases where a lexical, non-cliti c auxili ary, which according to

42

correct order within the ECC given the cliti cs-as-non-branching-elements hypothesis and the

economy of derivation condition that every requirement be satisfied through the shortest

movement possible.47 The gist of the analysis is the following: The verb moves up through

empty heads in (88). The cliti c-verb cluster is formed by left adjoining the cliti cs to the verb

instead of right adjoining the verb to the cliti cs, in accordance with Kayne’s LCA. Economy of

derivation ensures that the order of adjunctions is accusative-dative-auxiliary-negative cliti c.

(The relevant movements are shortest if they take place as soon as possible, that is, as soon as the

verb moves to a position c-commanding a cliti c.) Since the adjunctions take place to the left we

obtain the desired word order.

Let us spell -out the details of the analysis with respect to the derivation in (89). (I ignore

traces of cliti cs.)

(89) a. Ti [NegP ne [Neg’ [AuxP si [Aux’ [VP mu [V’ dal gi]]]] ]]

b. Ti [NegP ne [Neg’ [AuxP si [Aux’ [VP mu [V’ gi+dal]]] ]]]

c. Ti [NegP ne [Neg’ [AuxP si [Aux’ [gi+dal] i [VP mu [V’ ti]]]] ]]

d. Ti [NegP ne [Neg’ [AuxP si [Aux’ mu+[gi+dal] i [VP [V’ ti]]] ]]]

e. Ti [NegP ne [Neg’ [mu+[gi+dal] i] j [AuxP si [Aux’ tj [VP [V’ ti]]]] ]]

f. Ti [NegP ne [Neg’ si+[mu+[gi+dal] i ] j [AuxP [Aux’ tj [VP [V’ ti]]] ]]]

g. Ti [si+[mu+[gi+dal] i ] j]k [NegP ne [Neg’ tk [AuxP [Aux’ tj [VP [V’ ti]]] ]]]

h. Ti ne+[si+[mu+[gi+dal] i ] j]k [NegP [Neg’ tk [AuxP [Aux’ tj [VP [V’ ti]] ]]] ]

Krapova (1999) is categorially non-distinct from a verb, syntactically hosts a cliti c, as in (i),

under the plausible assumption that the cliti c is generated below the auxil iary.

(i) Petko go à á â á ã ä å æ á ç è Petko it was read

‘Petko had read it.’

47The requirement is responsible for Superiority effects. For example, given the structure in (ia)

prior to wh-movement, the requirement in question favors the movement of the first wh-phrase

to SpecCP over the movement of the second wh-phrase. The strong +wh-feature of C is checked

through shorter movement in (ib) than in (ic).

(i) a. +wh C John tell who that Mary should buy what

b. Who did John tell t that Mary should buy what?

c. *What did John tell who that Mary should buy t?

43

Assuming a c-command requirement on overt movement, the dative cliti c cannot incorporate

into the verb until the verb moves outside of the VP since the verb does not c-command the

dative cliti c in its base-generated position. On the other hand, the accusative cliti c can

incorporate into the verb either before or after V-movement. Notice, however, that the

incorporation results in shorter movement if it takes place before the verb moves. Given the

economy of derivation condition that every requirement be satisfied through the shortest

movement possible, the accusative cliti c then has to incorporate into the verb by left-adjoining

to it before the verb moves. The dative cliti c has to wait for the verb or, more precisely, the

accusative cliti c+verb complex, to move to a higher head position and then undergoes

incorporation into it through left-adjunction.48 We derive the correct order dative cliti c-

accusative cliti c-verb.

We see here a very interesting consequence of economy of derivation, which requires that

every syntactic requirement be satisfied through the shortest movement possible. Economy of

derivation imposes a sort of earliness requirement on the movement of X to Y if Y is to undergo

further movement to Z. X must move to Y as soon as possible; in particular, before Y moves to é ê ë ì í î ï í î ð ð ñ ò ï ó ô ð õ í ö ÷ ø ù õ ú ù û ü ý õ ð ð þ í ÿ ú í � ù � � � � � ò � � � �-156.)49

The rest of the derivation in (89) is straightforward. The auxili ary and the negative cliti c

have to wait for the dat+acc+V complex to move to a position c-commanding them in order to

undergo incorporation into the complex through left-adjunction. Economy of derivation forces

the following order of incorporation: 1. auxili ary cliti c 2. negative cliti c, since this way the

incorporation results in the shortest movements possible. We then obtain the desired word order

48If multiple adjunction to the same head is not allowed, as argued by Kayne (1994), the dative

cliti c would actually left-adjoin to the accusative cliti c, which is itself left-adjoined to the verb.

Notice that Kayne (1994) suggests that cliti cs do not adjoin to the finite verb. One could,

however, quite easily make room for such adjunction to take place in Bulgarian and Macedonian,

which seems necessary on empirical grounds, while still maintaining the gist of Kayne’s system.

(Kayne’s suggestion was made based on certain assumptions concerning the LCA and the sub-

word level structure that do not seem necessary.)

49Richard Kayne (personal communication) observes that the desired result can also be achieved

by appealing directly to Pesetsky’s (1989) Earliness Principle. Adopting Boškovi�’s (1998)

version of Chomsky’s (1995) definition of strong features (i.e. features that drive overt

movement), according to which strong features must be checked as soon as possible, would also

have the desired result.

44

negative cliti c-auxili ary cliti c-dative cliti c-accusative cliti c-verb.50

I conclude, therefore, that we can account for word order within the ECC in Bulgarian

and Macedonian (more precisely, the fact that the structural height of relevant elements prior to

the ECC formation corresponds to the left-to-right order within the ECC) without employing

rightward adjunction.

The above analysis of (86) is essentially forced on us by economy of derivation, the

cliti cs-as-non-branching-elements hypothesis, and the LCA. All the crucial ingredients of the

analysis are forced, or, more appropriately, provided for free, by one of these three mechanisms.

The cliti cs-as-non-branching-elements hypothesis forces generation of cliti cs in Spec positions,

economy of derivation imposes a particular ordering of cliti c adjunctions, and the LCA forces

the adjunctions to proceed to the left. The fact that the mechanisms in question conspire to force

an analysis that turns out to give us exactly what we need empirically provides strong evidence

for the mechanisms involved. It is also worth emphasizing that the current analysis achieves a � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � ! " " # � � � li ).

As Steven Franks (personal communication) observes, under the current analysis, which

combines the X0 and XP analyses of cliti cs, cliti cs are generated as morphological heads in

syntactic phrase positions. The fact that they are morphological heads makes it possible for them

to undergo head movement.

To summarize, in contrast to SC, Bulgarian and Macedonian have ECC formation (i.e.

cliti cs and the verb form a complex head in these languages). Contrary to standard assumptions,

the order within the ECC can be accounted for without assuming rightward adjunction given that

cliti cs move to the verb instead of the verb moving to cliti cs, as standardly assumed.

References

Abels, Klaus. (2001a) “Cliti cs as complements of P” . What Serbo-Croatian tells us about

preposition stranding. Paper presented at the Acme Balkanica Conference, Concordia

University, Montreal.

Abels, Klaus. (2001b) “* [P cliti c]! - Why?”. Paper presented at the 4th European Conference on

Formal Description of Slavic Languages, University of Potsdam.

Alexander, Ronelle. (1994) “The balkanization of Wackernagel’s Law”. Proceedings of the 9th

Biennial Conference on Balkan and South Slavic Linguistics, Literature, and Folklore

(=Indiana Linguistic Studies 7). Indiana University Linguistic Club, Bloomington, 1-8.

50The reader is referred to Boškovi (2001a) for demonstration how the is-final effect, which

Bulgarian also has, is captured in the current system.

45

Alexandrova, Galia.(1997) “Pronominal cliti cs as G(eneralized) F(amiliarity)-Licensing Agr0”.

Wayles Browne, Ewa Dornisch, Natasha Kondrashova, and Draga Zec, eds., Formal

Approaches to Slavic Linguistics: The Cornell Meeting 1995. Ann Arbor: Michigan

Slavic Publications, 1-31.

Anderson, Stephen. (1993) “Wackernagel's revenge: cliti cs, morphology, and the syntax of

second position” . Language 69: 68-98.

Anderson, Stephen. (1996) “How to put your clitics in their place or why the best account of

second position-phenomena may be a nearly optimal one”. The Linguistic Review 13:

165-191.

Avgustinova, Tania. (1994) “On Bulgarian verbal cliti cs” . Journal of Slavic Linguistics 2: 29-47.

Baker, Mark. 1988. Incorporation: A theory of grammatical function. Chicago: University of

Chicago Press.

Bennett, David. (1986) “Towards an explanation of word-order differences between Slovene and

Serbo-Croat” . The Slavonic and East European Review 64: 1-24.

Bennett, David. (1987) “Word order change in progress: The case of Slovene and Serbo-Croat

and its relevance for Germanic”. Journal of Linguistics 23: 269-287.

Berent, Gerald P. (1980) “On the realization of trace: Macedonian cliti c pronouns” . Catherine V.

Chvany and Richard D. Brecht, eds., Morphosyntax in Slavic. Columbus, Ohio: Slavica,

150-186.

Billi ngs, Loren, and Catherine Rudin. (1996) “Optimali ty and Superiority: A new approach to

overt multiple wh-ordering” . Jind$ ich Toman, ed., Formal Approaches to Slavic

Linguistics: The College Park Meeting, 1994. Ann Arbor: Michigan Slavic Publications,

35-60.

Bobalji k, Jonathan. (1995) Morphosyntax: The syntax of verbal inflection. Ph.D. dissertation,

MIT, Cambridge, Mass. % & ' ( ) * + , - . / 0 ( 1 2 . 3 1 2 . / 1 3 4 5 ' 6 1 2 & 7 0 8 9 : ; < < < = > ? @ ' ( A 0 4 0 ( B C @-connection: Evidence for

unselective attraction” . Tracy Holloway King and Irina Sekerina, eds., Formal

Approaches to Slavic Linguistics: The Philadelphia Meeting, 1999. Ann Arbor: Michigan

Slavic Publications, 22-40. % & D ) & 7 0 8 + E ' A 5 ) & 9 : F G G H = > I 1 / 4 0 ( 0 6 A ' J & 7 ' J ' 2 4 1 2 . K ' ( & 2 . 6 & K 0 4 0 & 2 ( A 0 4 0 ( 0 L 1 4 0 & 2 0 2Serbo-Croatian” . Lingua 96: 245-266. % & D ) & 7 0 8 + E ' A 5 ) & 9 : F G G M 1 =

The syntax of nonfinite complementation: An economy approach.

Cambridge, Mass.: MIT Press. % & D ) & 7 0 8 + E ' A 5 ) & 9 : F G G M N = > O / & 2 4 0 2 Pwh-phrases in Serbo-Croatian” . Martina Lindseth and

Steven Franks, eds., Formal Approaches to Slavic Linguistics: The Indiana Meeting,

1996. Ann Arbor: Michigan Slavic Publications, 86-107.

46

Q R S T R U V W X Y Z [ \ T R ] ^ _ ` ` a b c d e [ f V g [ Z h i-fronting and economy of derivation” . Proceedings of the

West Coast Conference on Formal Linguistics 16. Stanford, Cali f.: CSLI Publications,

49-63. Q R S T R U V W X Y Z [ \ T R. (1999) “On multiple feature-checking: Multiple wh-fronting and multiple

head-movement” . Samuel Epstein and Norbert Hornstein, eds., Working Minimalism.

Cambridge, Mass: MIT Press, 159-187. Q R S T R U V W X Y Z [ \ T R ] ^ j k k k l b c m Z n R o p g R q V f V R o n [ V f V n V q l f V R o r m s o f l tand/or phonology?” Frits

Beukema and Marcel den Dikken, eds., Cliti c phenomena in European languages.

Amsterdam: John Benjamins, 71-119. Q R S T R U V W X Y Z [ \ T R ] ^ j k k k u b c v i l f V q q g Z n V l [ l u R e f w e [ f V g [ Z h i-fronting?” . In Proceedings of

NELS 30. GLSA, University of Massachusetts, Amherst, 83-107. Q R S T R U V W X Y Z [ \ T R ] ^ j k k _ l bOn the nature of the syntax-phonology interface: Cliti cization and

related phenomena. Amsterdam: Elsevier Science. Q R S T R U V W X Y Z [ \ T R ] ^ j k k _ u b “L i without PF movement” . Steven Franks, Tracy King, and Misha

Yadroff, eds., Formal Approaches to Slavic Linguistics: The Indiana Meeting, 2000. Ann

Arbor: Michigan Slavic Publications, 57-75. x y z { y | } ~ � � � � � { y � � � � � � � � � � � � � � � � � � � � � � � � � � � � } y � � � � � � � � � � � � � � � � � � � � � � � th European

Conference on Formal Description of Slavic Languages, University of Potsdam. x y z { y | } ~ � � � � � { y � � � � � � � � � � � } � } � � � � � y �-branching elements and the linear correspondence

axiom”. Linguistic Inquiry 33: 329-340. x y z { y | } ~ � � � � � { y � � � � � � � � � � � � � � � } � � � � �-fronting” . Linguistic Inquiry 33: 351-383. x y z { y | } ~ � � � � � { y � � � � � � � � � x � � � � � � � � � � � � � � y � � � y � � � y � �   � � � � } � } � � � � � ¡ � � � ¢ � } | � � � } � � y �

Connecticut, Storrs. x y z { y | } ~ � � � � � { y � � � � £ � � | � � ¤ � � � { � � � � � � � � � � � y � y � y ¥ �-syntax interactions in South Slavic”.

Balkanistica 15: 49-74.

Browne, Wayles. (1974) “On the problem of encliti c placement in Serbo-Croatian” . Richard D.

Brecht and Catherine V. Chvany, eds., Slavic transformational syntax. Ann Arbor:

University of Michigan, 36-52.

Browne, Wayles. (1975) “Serbo-Croatian encliti cs for English-speaking learners” . Rudolf ¤ } � } � y | } ~ � � � �Contrastive analysis of English and Serbo-Croatian, Vol. 1. Zagreb:

Institute of Linguistics, Faculty of Philosophy, University of Zagreb, 105-134.

Browne, Wayles. (1986) “Parameters in cliti c placement” . Paper presented at the 5th Biennial

Conference on Balkan and South Slavic Linguistics, Literature, and Folklore, Indiana

University.

Browne, Wayles. (1993) “Serbo-Croat” . Bernard Comrie and Grevil le Corbett, eds., The

Slavonic languages. London: Routledge, 306-387.

47

Browne, Wayles. (1994) “Cliti c placement in Slovenian subordinate clauses” . Paper presented at

the 9th Biennial Conference on Balkan and South Slavic Linguistics, Literature, and

Folklore, Indiana University.

Caink, Andrew. (1998) The Lexical interface: Closed class items in South Slavic and English.

Ph.D. dissertation, University of Durham.

Caink, Andrew. (1999) “Serbian/Croatian/Bosnian cliti cs at the lexical interface”. Katarzyna

Dziwirek, Herbert Coats, Cynthia M. Vakerliyska, eds., Formal Approaches to Slavic

Linguistics: The Seattle Meeting 1998. Ann Arbor: Michigan Slavic Publications, 81-100.

Cardinaletti, Anna and Michal Starke (1999) “The typology of structural deficiency: A case

study of the three classes of pronouns” . Hank van Riemsdijk, ed., Clitics in the languages

of Europe. Berlin: Mouton, 185-234. ¦ § ¨ © ª « ¬ ­ ª ® ¯ § ° ± ² ³ ³ ´ µ ¶ · ¸ « ¹ © º » ¼ ® ½ º § ª ª ½ § ¾ ½ ¹ ¼ ¸ « ¿ ¸ ½ À ¼ Á À ½ º ½ Â ® º § ª ¹ ½ À Â ¿ § º ¾ ¼ ¸ « ¾ ½ © Ã ª ® º Ä ½ ¹ ¼ ¸ «object in Macedonian” . Journal of Slavic Linguistics 5: 3-19.

Å avar, Damir, and Chris Wilder. (1994) “Cliti c third in Croatian” . Eurotyp Working Papers,

Grammatical Models Selection 6, 19-61.

Chomsky, Noam. (1994) “Bare phrase structure”. MIT Occasional Papers in Linguistics 5.

MITWPL, Department of Linguistics and Philosophy, MIT, Cambridge, Mass.

Chomsky, Noam. (1995) “Categories and transformations” . The Minimalist program.

Cambridge, Mass.: MIT Press.

Cinque, Guglielmo. (1998) Adverbs and functional heads: A cross-linguistic perspective.

Oxford: Oxford University Press.

Comrie, Bernard. (1981) Language universals and linguistic typology: Syntax and morphology.

Chicago: University of Chicago Press.

Derbyshire, Willi am. (1993) A basic reference grammar of Slovene. Columbus, Ohio: Slavica

Publishers.

Dikken, Marcel den. (1994) “Auxili aries and participles” . Proceedings of NELS 24. GLSA,

University of Massachusetts, Amherst, 65-79.

Dimitrova-Vulchanova, Mila. (1995) “Cliti cs in Slavic”. Studia Linguistica 49: 54-92.

Dimitrova-Vulchanova, Mila, and Lars Hellan. (1999) “Cliti cs and Bulgarian clause structure”.

Henk van Riemsdijk, ed., Cliti cs in the languages of Europe. Berlin: Mouton de Gruyter,

469-514.

Embick, David, and Roumyana Izvorski. (1997) “Participle-auxili ary word orders in Slavic”.

Wayles Browne, Ewa Dornisch, Natasha Kondrashova, and Draga Zec, eds., Formal

Approaches to Slavic Linguistics: The Cornell Meeting 1995. Ann Arbor: Michigan

Slavic Publications, 210-239.

Ewen, Robert. (1979) A grammar of Bulgarian cliti cs. Ph.D. dissertation, University of

48

Washington, Seattle.

Franks, Steven. (1997) “South Slavic cli tic placement is still syntactic”. Penn Working Papers in

Linguistics 4: Proceedings of the 21st Annual Penn Linguistics Colloquium. Department

of Linguistics, University of Pennsylvania, Philadelphia, 111-126.

Franks, Steven. (1998) “Cliti cs in Slavic”. Presented at the Comparative Slavic Morphosyntax

Workshop, Bloomington, Indiana.

Franks, Steven. (1999) “Optimali ty Theory and clit ics at PF”. Katarzyna Dziwirek, Herbert

Coats, Cynthia M. Vakerliyska, eds., Formal Approaches to Slavic Linguistics: The

Seattle Meeting 1998. Ann Arbor: Michigan Slavic Publications, 101-116.

Franks, Steven. (2000a) “Cliti cs at the interface: An introduction to Cli tic phenomena in

European languages” . Frits Beukema and Marcel den Dikken, eds., Cliti c phenomena in

European languages. Amsterdam: John Benjamins, 1-46.

Franks, Steven. (2000b) “The internal structure of Slavic Noun Phrases, with special reference to

Bulgarian” . Paper presented at Generative Linguistics in Poland 2, University of Warsaw,

Warsawa.

Franks, Steven, and Tracy King. (2000) A handbook of Slavic cliti cs. Oxford: Oxford University

Press.

Franks, Steven, and Lji ljana Progovac. (1994) “On the placement of Serbo-Croatian cliti cs” .

Proceedings of the 9th Biennial Conference on Balkan and South Slavic Linguistics,

Literature, and Folklore (=Indiana Linguistic Studies 7). Indiana University Linguistic

Club, Bloomington, 69-78.

Fried, Mirjam. (1999) “The “ free” datives in Czech as a linking problem”. Katarzyna Dziwirek,

Herbert Coats, Cynthia M. Vakerliyska, eds., Formal Approaches to Slavic Linguistics:

The Seattle Meeting 1998. Ann Arbor: Michigan Slavic Publications, 145-166.

Golden, MaÆ Ç È É Ê É Ë Ì Í Ç Î Ï Ë É Í Ç Î Ð È Ï Ñ Ç Ò Ó Ô Ï Õ Õ É Æ Ì Ö × Ø Ù Ù Ù Ú Û Ó Î Ð Ñ Ï Ë Ï Õ Æ Ð Ë Ð Ü Ç Ë É Î Ý Î Ç Þ Ç Ý ß à Ö á Æ Ç Þ ßBeukema and Marcel den Dikken, eds., Cliti c phenomena in European languages.

Amsterdam: John Benjamins, 191-207.

Halpern, Aaron Lars. (1995) On the placement and morphology of cliti cs. Stanford, Cali f.: CSLI

Publications.

Hayes, Bruce. (1989) “The prosodic hierarchy in meter” . Paul Kiparsky and Gilbert Youmans,

eds., Rhythm and meter. Phonetics and phonology, Vol. 1. San Diego: Academic Press,

201-260.

Hiramatsu, Kazuko. (2000a) Accessing linguistic competence: Evidence from children’s and

adults’ acceptabilit y judgments. Ph.D. dissertation, University of Connecticut, Storrs.

Hiramatsu, Kazuko. 2000b. “What did Move didn’ t erase?” Dave Braze, Yutaka Kudo, and

Kazuko Hiramatsu, eds., University of Connecticut Working Papers in Linguistics 10.

49

Department of Linguistics, University of Connecticut, Storrs, 75-110. [Distributed by

MITWPL, Department of Linguistics and Philosophy, MIT, Cambridge, Mass.]

Hock, Hans Heinrich. (1992) “What's a nice word like you doing in a place like this” . Studies in

Linguistic Sciences 22: 39-87.

Izvorski, Roumyana, Tracy Holloway King, and Catherine Rudin. (1997) “Against Li lowering

in Bulgarian” . Lingua 102: 187-194.

Kayne, Richard. (1994) The antisymmetry of syntax. Cambridge, Mass.: MIT Press.

King, Tracy. (1996) “Slavic cliti cs, long head movement, and prosodic inversion” . Journal of

Slavic Linguistics 4: 274-311.

Kiss, Katalin. (1995) Discourse configurational languages. New York: Oxford University Press.

Krapova, Iliyana. (1997) “Auxili aries and complex tenses in Bulgarian” . Wayles Browne, Ewa

Dornisch, Natasha Kondrashova, and Draga Zec, eds., Formal Approaches to Slavic

Linguistics: The Cornell Meeting 1995. Ann Arbor: Michigan Slavic Publications, 320-

344.

Krapova, Iliyana. (1999) “The system of auxili aries in Bulgarian” . Mila Dimitrova-Vulchanova

and Lars Hellan, eds., Topics in South Slavic Syntax and Semantics. Amsterdam: John

Benjamins, 59-89.

Lambova, Mariana. (2002) “On A’ -movements in Bulgarian and their interaction” . The

Linguistic Review 18: 327-374.

Larson, Richard. (1988) “On the double object construction” . Linguistic Inquiry 19: 335-391.

Lasnik, Howard. (1995) “A note on pseudogapping” . Rob Pensalfini and Hiroyuki Ura, eds.,

Papers on minimalist syntax, MIT working papers in linguistics 27. MITWPL,

Department of Linguistics and Philosophy, MIT, Cambridge, Mass., 143-163.

Law, Paul. (2001) “A prosodic account for cliti c placement in Serbo-Croatian” . Paper presented

at the 4th European Conference on Formal Description of Slavic Languages, University of

Potsdam.

Legendre, Géraldine. (1998) “Second position cliti cs in a verb-second language: Conflict

resolution in Macedonian” . Proceedings of ESCOL 14. GLC Publications, Cornell

University, Ithaca, NY, 139-149.

Legendre, Géraldine. (1999) “Morphological and prosodic alignment at work: The case of South-

Slavic cliti cs” . Proceedings of the West Coast Conference on Formal Linguistics 17.

Stanford, Cali f.: CSLI Publications, 436-450.

Legendre, Géraldine. (2000) “Morphological and prosodic alignment of Bulgarian cliti cs” . Joost

Dekkers, Frank van der Leeuw, and Jeroen van der Weijer, eds., Optimality theory:

Syntax, phonology, and acquisition. Oxford: Oxford University Press, 423-462.

Marantz, Alec. (1988) “Cliti cs, morphological merger, and the mapping to phonological

50

structure”. Michael Hammond and Michael Noonan, eds., Theoretical morphology:

Approaches in modern linguistics. San Diego: Academic Press, 253-270.

Marantz, Alec. (1989) “Cliti cs and phrase structure”. Mark Baltin and Anthony Kroch, eds.,

Alternative Conceptions of Phrase Structure. Chicago: University of Chicago Press, 99-

116.

Marantz, Alec. (1993) “ Implications and asymmetries in double object constructions” . Sam A.

Mchombo, ed., Theoretical aspects of Bantu grammar. Stanford, Cali f.: CLSI

Publications, 113-150.

Miyagawa, Shigeru. (1997) “Against optional scrambling” . Linguistic Inquiry 28: 1-25.

Nespor, Marina, and Irene Vogel. (1982) “Prosodic domains of external sandhi rules” . Harry

van der Hulst and Norval Smith, eds., The structure of phonological representation (Part

1). Dordrecht: Foris, 225-255.

Nespor, Marina, and Irene Vogel. (1986) Prosodic phonology. Dordrecht: Foris.

Orešnik, Janez. (1983-1984) “Slovenske breznaglasnice se vedejo predusem kot prokliti ke”.

Jezik in Slovstvo 29: 129. â ã ä å ã æ ç è é ê ë ì ä í î ï ð ð ñ òBó lgarski sintaksis. Upravlenie i svó rzvane. Plovdiv: Plovdivsko

Universitetsko Izdatelstvo.

Percus, Orin. (1993) “The captious cliti c: Problems in Serbo-Croatian cliti c placement” . Ms.,

MIT, Cambridge, Mass.

Pesetsky, David. (1989) “The Earliness principle” . Ms., MIT, Cambridge, Mass.

Pesetsky, David. (1997). “Optimali ty theory and syntax: Movement and pronunciation” . Diana

Archangeli and D. Terence Langendoen, eds., Optimality theory: An overview. Malden,

Mass.: Blackwell , 134-170.

Pesetsky, David. (1998) “Some optimali ty principles of sentence pronunciation” . Pilar Barbosa,

Danny Fox, Paul Hagstrom, Martha McGinnis, and David Pesetsky, eds., Is the Best

Good Enough? Cambridge, Mass.: MIT Press and MIT WPL, Department of Linguistics

and Philosophy, MIT, Cambridge, Mass, 337-383.

Philli ps, Colin. (1996) Order and structure. Ph.D. dissertation, MIT, Cambridge, Mass.

Priestly, Tom. (1996) “Slovene”. Bernard Comrie and Grevill e G. Corbett, eds., The Slavonic

Languages. London: Routledge, 388-451.

Progovac, Ljilj ana. (1993) “Locali ty and subjunctive-like complements in Serbo-Croatian” .

Journal of Slavic Linguistics 1: 116-144.

Progovac, Ljilj ana. (1996) “Cliti cs in Serbian/Croatian: Comp as the second position” . Aaron

Halpern and Arnold Zwicky, eds., Approaching second: Second position cliti cs and

related phenomena. Stanford, Cali f..: CSLI Publications, 411-428.

Progovac, Ljilj ana. (1998). “Cliti c clusters and coordination” . Proceedings of ESCOL 14. GLC

51

Publications, Cornell University, Ithaca, NY, 161-169.

Progovac, Ljilj ana. (1999) “Eventive to and the placement of cliti cs” . Istvan Kenesei, ed.,

Crossing boundaries: Advances in the theory of central and eastern European languages.

Amsterdam: John Benjamins, 33-44.

Progovac, Ljilj ana. (2000) “Where do cliti cs cluster?” Frits Beukema and Marcel den Dikken,

eds., Cliti c phenomena in European languages. Amsterdam: John Benjamins, 249-258. ô õ ö õ ÷ ø ù ú û-ü ø ý ú û þ ÿ � � ÷ õ � � � � � � �

The grammar of Serbo-Croatian cliti cs: A synchronic and

diachronic perspective. Ph.D. dissertation, University of Illi nois, Urbana-Champaign. ô õ ö õ ÷ ø ù ú û-ü ø ý ú û þ ÿ � � ÷ õ � � � � � � � � õ ý � � � ÷ ø � � � � � ø

-Croatian cliti cs: A prosodic approach” .

Aaron Halpern and Arnold Zwicky, eds., Approaching second: Second position cliti cs

and related phenomena. Stanford, Cali f.: CSLI Publications, 429-445.

Rå Hauge, Kjetil . (1976) The word order of predicate cliti cs in Bulgarian. Meddelelser 10,

Slavisk-Baltisk Institutt, Universitetet i Oslo. [Republished as Rå Hauge, Kjetil . 1999.

Journal of Slavic Linguistics 7: 89-137.]

Rivero, María Luisa. (1993) “Bulgarian and Serbo-Croatian yes-no questions: V raising to li

versus li hopping” . Linguistic Inquiry 24: 567-575.

Rivero, María Luisa. (1997) “On two locations for complement cliti c pronouns: Serbo-Croatian,

Bulgarian, and Old Spanish” . Ans van Kemenade and Nigel Vincent, eds., Parameters of

morphosyntactic change. Cambridge: Cambridge University Press, 170-206.

Roberts, Ian. (1994) “Second position effects and agreement in Comp”. Paper presented at the

Third Annual Workshop on Formal Approaches to Slavic Linguistics, University of

Maryland, College Park.

Rudin, Catherine. (1988) “On multiple questions and multiple wh-fronting” . Natural Language

and Linguistic Theory 6: 445-501.

Rudin, Catherine. (1993) “On focus position and focus marking in Bulgarian questions” . Papers

from the 4th Annual Meeting of the Formal Linguistics Society of Midamerica. University

of Iowa, Iowa City, 252-265.

Rudin, Catherine. (1997) “AgrO and Bulgarian pronominal cliti cs” . Martina Lindseth and Steven

Franks, eds., Formal Approaches to Slavic Linguistics: The Indiana Meeting, 1996. An

Arbor: Michigan Slavic Publications, 224-252.

Rudin, C., T. Holloway King, and R. Izvorski (1998) “Focus in Bulgarian and Russian Yes/No

questions” . Elena Benedicto, Mirabel Romero, and Sathoshi Tomioka, eds., Proceedings

of the Workshop on Focus, University of Massachusetts Occasional Papers in Linguistics

21. GLSA, University of Massachusetts, Amherst, 209-226.

Rudin, Catherine, Christina Kramer, Loren Billi ngs, and Matthew Baerman. (1999)

“Macedonian and Bulgarian li questions: Beyond syntax” . Natural Language and

52

Linguistic Theory 17: 541-586.

Sadock, Jerrold. (1991) Autolexical syntax. Chicago: University of Chicago Press.

Schick, Ivanka. (2000) “Cliti c doubling constructions in Balkan-Slavic languages” . Frits

Beukema and Marcel den Dikken, eds., Cliti c phenomena in European languages.

Amsterdam: John Benjamins, 259-292.

Schütze, Carson. (1994) “Serbo-Croatian second position cliti c placement and the phonology-

syntax interface” . MIT Working Papers in Linguistics 21. MITWPL, Department of

Linguistics and Philosophy, MIT, Cambridge, Mass., 373-473.

Selkirk, Elisabeth. (1986) Phonology and syntax: The relation between sound and structure.

Cambridge, Mass.: MIT Press.

Sproat, Richard. (1988) “Bracketing paradoxes, cliti cization, and other topics: The mapping

between syntactic and phonological structure”. Martin Everaert, Arnold Evers, Riny

Huybregts, and Mieke Trommelen, eds., Morphology and modularity. Dordrecht: Foris,

339-360. � � � � � � � � � � � � � � � � � � ! " # # $ % & � ' � � �-distance movement of wh-phrases in Serbo-Croatian matrix

clauses” . Ms., University of Connecticut, Storrs. � � � � � � � � � � � � � � � � � � ! " # # ( � % & ) � � ' � � * � + � , � � � � - � � � . �-Croatian cliti cs: Evidence from VP

elli psis” . Linguistic Inquiry 29: 527-537. � � � � � � � � � � � � � � � � � � ! " # # ( . % & ) � � ' � � * � + � , � � � �f Serbo-Croatian cliti cs: Evidence from cliti c + * � , . � � / � � � 0 1 � * * � � 2 � 2 3 4 � * � 5 � 6 � 7 5 � � � � � � � � � � � 8 � � � 5 2 � � � � 9 � * * � � , � � : � � � � � � 2 �

Formal Approaches to Slavic Linguistics: The Connecticut Meeting, 1997. Ann Arbor:

Michigan Slavic Publications, 267-286.

S� � � � � � � � � � � � � � � � � ! " # # ( + % & 0 1 � * * � � 2 � 2 � � � 0 � � � 2 � � / * � � / ; � / � < � , � * � + � � � � � 2 - � � � � � . � *

morphology. In Proceedings of ESCOL 14. GLC Publications, Cornell University, Ithaca,

NY, 192-203. � � � � � � � � � � � � � � � � � � ! " # # # %What do second position cliti cization, scrambling, and multiple wh-

fronting have in common. Ph.D. dissertation, University of Connecticut, Storrs.

Taylor, Ann. (1990) Cliti cs and configurationality in Ancient Greek. Ph.D. dissertation,

University of Pennsylvania, Philadelphia. = � , � � � > � � � ? � + ' (1999) “On cliti c displacement” . Mila Dimitrova-Vulchanova and Lars Hellan,

eds., Topics in South Slavic Syntax and Semantics. Amsterdam: John Benjamins, 205-

228. = � , � � � ) * / � @ � 7 � 2 5 � ! " # # A % & = ' � 6 � * 5 � � � * � � � + + * � ; 2 � * + * � � � + 2 3 Natural Language and

Linguistic Theory 14: 811-872. = � , � � � ) * / � @ � 7 � 2 5 � ! " # # B % & C � �-first as a default cliti c position” . Journal of Slavic

Linguistics 5: 1-23.

53

D E F G H I J K L M N G O P Q R M S T U V V V W X J Y Z K G [ G Z Q G [ P Q \ S ] ^ G [ Q _ P ` R P F M M Y a N M ^ Z P K a P Y b G R R P Y I P a Q S ICliti c phenomena in European languages. Amsterdam: John Benjamins, 293-316.

Toporiši c I d E e P S T f g h i W Slovenska slovnica. Maribor: Obzorja.

Watanabe, Akira. (1993) AGR-based Case theory and its interaction with the A'-system. Ph.D.

dissertation, MIT, Cambridge, Mass.

Wilder, Chris. (1997) “Some properties of elli psis in coordination” . Artemis Alexiadou and T.

Alan Hall , eds., Studies in universal grammar and typological variation. Amsterdam:

John Benjamins, 59-107. j G K a P ^ I k l ^ G Q I M Y a b M F G ^ m M n M ^ S T f g g i M W X o E Y L l P M a F E n ement? Verb movement and

cliti cization in Croatian” . Lingua 93: 1-58. j G K a P ^ I k l ^ G Q I M Y a b M F G ^ m M n M ^ S T f g g i p W X j E ^ a E ^ a P ^ n M ^ G M [ G E Y I n P ^ p F E n P F P Y [ I M Y a P Z E Y E F qprinciples” . Studia Linguistica 48: 46-86. j G K a P ^ I k l ^ G Q M Y a b M F G ^ m M n M ^ S T f g g r W X s P ^ p F E n P F P Y [ , cliti cization, and coordination” . Ms.,

ZAS, Berlin and Universität Hamburg.

Zec, Draga, and Sharon Inkelas. (1990) “Prosodically constrained syntax” . Sharon Inkelas and

Draga Zec, eds., The Phonology-Syntax Connection. Chicago, Ill .: University of Chicago

Press, 365-378.

Department of Linguistics, U-1145

University of Connecticut

Storrs, CT 06269

boskovic@sp.uconn.edu